DISQLite3: Version History

Update to SQLite Version 3.47.0.

• Allow arbitrary expressions in the second argument to the RAISE function.
• If the RHS of the →> operator is negative, then access array elements counting from the right.
• FTS5 tables can now be dropped even if they use a non-standard tokenizer that has not been registered.
• Fix the group_concat() aggregate function so that it returns an empty string, not a NULL, if it receives a single input value which is an empty string.
• Enhance the generate_series table-valued function so that it is able to recognize and use constraints on its output value.
• Preupdate hooks now recognize when a column added by ALTER TABLE ADD COLUMN has a non-NULL default value.
• Enhance the sqlite_dbpage table-valued function such that INSERT can be used to increase or decrease the size of the database file.

Performance optimizations:

• Improved reuse of subqueries associated with the IN operator, especially when the IN operator has been duplicated due to predicate push-down.
• Use a Bloom filter on subqueries on the right-hand side of the IN operator, in cases where that seems likely to improve performance.
• Ensure that queries like “SELECT func(a) FROM tab GROUP BY 1” only invoke the func() function once per row.
• No attempt is made to create automatic indexes on a column that is known to be non-selective because of its use in other indexes that have been analyzed.
• Adjustments to the query planner so that it produces better plans for star queries with a large number of dimension tables.
• Add the “order-by-subquery” optimization, that seeks to disable sort operations in outer queries if the desired order is obtained naturally due to ORDER BY clauses in subqueries.
• The “indexed-subtype-expr” optimization strives to use expressions that are part of an index rather than recomputing the expression based on table values, as long as the query planner can prove that the subtype of the expression will never be used.

FTS5 Changes:

• Add the fts5_tokenizer_v2 API and the locale=1 option, for creating custom locale-aware tokenizers and fts5 tables that may take advantage of them. See DISQLite3Fts5Tokenizer_v2 for a fts5_tokenizer_v2 example tokenizer.
• Add the contentless_unindexed=1 option, for creating contentless fts5 tables that store the values of any UNINDEXED columns persistently in the database.
• Allow an FTS5 table to be dropped even if it uses a custom tokenizer whose implementation is not available.

Update to SQLite Version 3.46.1.

• Improved robustness while parsing the tokenize= arguments in FTS5.
• Enhancements to covering index prediction in the query planner. Add early detection of over-prediction of covering indexes so that sqlite3_prepare will return an error rather than just generate bad bytecode.
• Do not let the number of terms on a VALUES clause be limited by SQLITE_LIMIT_COMPOUND_SELECT, even if the VALUES clause contains elements that appear to be variables due to double-quoted string literals.
• Fix the window function version of lang_aggfunc.html#group_concat so that it returns an empty string if it has one or more empty string inputs.
• In FTS5 secure-delete mode, fix false-positive integrity-check reports about corrupt indexes.
• Syntax errors in ALTER TABLE should always return SQLITE_ERROR. In some cases, they were formerly returning SQLITE_INTERNAL.
• Other minor fixes.

Update to SQLite Version 3.46.0.

• Enhance PRAGMA optimize in multiple ways, to make it simpler to use:
  • PRAGMA optimize automatically implements a temporary analysis limit to prevent excess runtime on large databases.
  • Added the new 0x10000 bitmask option to check for updates on all tables.
  • Automatically re-analyze tables that do not have sqlite_stat1 entries.
• Enhancements to the date and time functions:
  • The strftime() SQL function now supports %G, %g, %U, and %V.
  • New modifiers 'ceiling' and 'floor' control the algorithm used to resolve ambiguous dates when shifting a date by an integer number of months and/or years.
  • The 'utc' and 'localtime' modifiers are now no-ops if SQLite knows that the time is already in UTC or in the localtime, respectively.
• Add support for underscore (“_”) characters between digits in numeric literals.
• Add the json_pretty SQL function.
• Query planner improvements:
  • The “VALUES-as-coroutine” optimization enables INSERT statements with thousands of rows in the VALUES clause to parse and run in about half the time and using about half as much memory.
  • Allow the use of an index for queries like “SELECT count(DISTINCT col) FROM …”, even if the index records are not smaller than the table records.
  • Improved recognition of cases where the value of an SQL function is constant because all its arguments are constant.
  • Enhance the WHERE-clause push-down optimization so that it is able to push down WHERE clause terms containing uncorrelated subqueries.
• Allocate additional memory from the heap for the SQL parser stack if that stack overflows, rather than reporting a “parser stack overflow” error.
• JSON changes:
  • Allow ASCII control characters within JSON5 string literals.
  • Fix the → and →> operators so that when the right-hand side operand is a string that looks like an integer it is still treated as a string, because that is what PostgreSQL does.
• Allow large hexadecimal literals to be used as the DEFAULT value to a table column.
• Add the randomjson extension.

• Update to SQLite Version 3.45.3.
  • Fix a long-standing bug that might (rarely) cause the “old.*” values of an UPDATE trigger to be incorrect if that trigger fires in response to an UPSERT.
  • Fix a bug in sum that could cause it to return NULL when it should return Infinity.
  • Enforce that application defined SQL functions that invoke sqlite3_result_subtype must be registered with SQLITE_RESULT_SUBTYPE. Failing to do so now causes sqlite3_result_subtype to raise an SQL error whereas before it might have behaved as a no-op.

• Update to SQLite Version 3.45.2.
  • Fix an error in UPSERT that could cause an index to get out-of-sync with its table.
  • Reduce the scope of the NOT NULL strength reduction optimization. The optimization was being attempted in some contexts where it did not work, resulting in incorrect query results.
  • Other trifling corrections and fixes that have come up since the previous patch release.

• Update to SQLite Version 3.45.1.
  • Restore the JSON BLOB input bug, and promise to support the anomaly in subsequent releases, for backward compatibility.
  • Fix the PRAGMA integrity_check command so that it works on read-only databases that contain FTS3 and FTS5 tables.
  • Fix issues associated with processing corrupt JSONB inputs:
    • Prevent exponential runtime when converting a corrupt JSONB into text.
    • Fix a possible read of one byte past the end of the JSONB blob when converting a corrupt JSONB into text.
  • Fix a long-standing bug in which a read of a few bytes past the end of a memory-mapped segment might occur when accessing a craftily corrupted database using memory-mapped database.
  • Fix a long-standing bug in which a nil pointer dereference might occur in the bytecode engine due to incorrect bytecode being generated for a class of SQL statements that are deliberately designed to stress the query planner but which are otherwise pointless.

• Update to SQLite Version 3.45.0.
  • Added the SQLITE_RESULT_SUBTYPE property for application-defined SQL functions. All application defined SQL functions that invokes sqlite3_result_subtype must be registered with this new property. Failure to do so might cause the call to sqlite3_result_subtype to behave as a no-op.
  • Enhancements to the JSON SQL functions:
    • All JSON functions are rewritten to use a new internal parse tree format called JSONB. The new parse-tree format is serializable and hence can be stored in the database to avoid unnecessary re-parsing whenever the JSON value is used.
    • New versions of JSON-generating functions generate binary JSONB instead of JSON text.
    • The json_valid() function adds an optional second argument that specifies what it means for the first argument to be “well-formed”.
  • Add the FTS5 tokendata option to the FTS5 virtual table.
  • Query planner improvements:
    • Do not allow the transitive constraint optimization to trick the query planner into using a range constraint when a better equality constraint is available.
    • The query planner now does a better job of disregarding indexes that ANALYZE identifies as low-quality.
  • Increase the default value for SQLITE_MAX_PAGE_COUNT from 1073741824 to 4294967294.

• Update to SQLite Version 3.44.2.
  • Fix a problem in FTS5 that was discovered during internal fuzz testing.
  • Other obscure bug fixes.

• Support Delphi 12 Athens Win32 and Win64.
• Update to SQLite Version 3.44.0.
  • Aggregate functions can now include an ORDER BY clause after their last parameter. The arguments to the function are processed in the order specified. This can be important for functions like string_agg and json_group_array.
  • Add support for the concat and concat_ws scalar SQL functions, compatible with PostgreSQL, SQLServer, and MySQL.
  • Add support for the string_agg aggregate SQL function, compatible with PostgreSQL and SQLServer.
  • New conversion letters on the strftime() SQL function: %e %F %I %k %l %p %P %R %T %u
  • Add new APIs: sqlite3_get_clientdata and sqlite3_set_clientdata.
  • Many errors associated with CREATE TABLE are now raised when the CREATE TABLE statement itself is run, rather than being deferred until the first time the table is actually used.
  • The PRAGMA integrity_check command now verifies the consistency of the content in various built-in virtual tables using the new xIntegrity method. This works for the FTS3, FTS4, FTS5, RTREE, and GEOPOLY extensions.
  • The SQLITE_DBCONFIG_DEFENSIVE setting now prevents PRAGMA writable_schema from being turned on. Previously writable_schema could be turned on, but would not actually allow the schema to be writable. Now it simply cannot be turned on.
  • Tag the built-in FTS3, FTS4, FTS5, RTREE, and GEOPOLY virtual tables as SQLITE_VTAB_INNOCUOUS so that they can be used inside of triggers in high-security deployments.
  • The PRAGMA case_sensitive_like statement is deprecated, as its use when the schema contains LIKE operators can lead to reports of database corruption by PRAGMA integrity_check.
  • Query planner optimizations:
    • In partial index scans, if the WHERE clause implies a constant value for a table column, replace occurrences of that table column with the constant. This increases the likelihood of the partial index being a covering index.
    • Disable the view-scan optimization as it was causing multiple performance regressions. In its place, reduce the estimated row count for DISTINCT subqueries by a factor of 8.

Update to SQLite Version 3.43.2.

• Fix obscure use after free (UAF) errors in JSON and FTS5.
• Fix an obscure memory leak in sqlite3_blob_reopen.
• Improve detection of corrupt databases.

Update to SQLite Version 3.43.1.

• Fix a regression in the way that the sum(), avg(), and total() aggregate functions handle infinities.
• Fix a bug in the json_array_length() function that occurs when the argument comes directly from json_remove.
• Fix the omit-unused-subquery-columns optimization so that it works correctly if the subquery is a compound where one arm is DISTINCT and the other is not.
• Other minor fixes.

Update to SQLite Version 3.43.0.

Key enhancements include added support for Contentless-Delete FTS5 Indexes, and performance improvements in JSON processing.

Details:

• Add support for Contentless-Delete FTS5 Indexes. This is a variety of FTS5 full-text search index that omits storing the content that is being indexed while also allowing records to be deleted.
• Enhancements to the date and time functions:
  • Added new time shift modifiers of the form ±YYYY-MM-DD HH:MM:SS.SSS.
  • Added the timediff() SQL function.
• Added the octet_length(X) SQL function.
• Added the sqlite3_stmt_explain API.
• Query planner enhancements:
  • Generalize the LEFT JOIN strength reduction optimization so that it works for RIGHT and FULL JOINs as well. Rename it to OUTER JOIN strength reduction.
  • Enhance the theorem prover in the OUTER JOIN strength reduction optimization so that it returns fewer false-negatives.
• Enhancements to the decimal extension:
  • New function decimal_pow2(N) returns the N-th power of 2 for integer N between -20000 and +20000.
  • New function decimal_exp(X) works like decimal(X) except that it returns the result in exponential notation - with a “e+NN” at the end.
  • If X is a floating-point value, then the decimal(X) function now does a full expansion of that value into its exact decimal equivalent.
• Performance enhancements to JSON processing results in a 2x performance improvement for some kinds of processing on large JSON strings.

• Add the FTS5 secure-delete command. This option causes all forensic traces to be removed from the FTS5 inverted index when content is deleted.
• Enhance the JSON SQL functions to support JSON5 extensions.
• The SQLITE_CONFIG_LOG and SQLITE_CONFIG_PCACHE_HDRSZ calls to sqlite3_config are now allowed to occur after sqlite3_initialize.
• New sqlite3_db_config options: SQLITE_DBCONFIG_STMT_SCANSTATUS and SQLITE_DBCONFIG_REVERSE_SCANORDER.
• Query planner improvements:
  • Enable the “count-of-view” optimization by default.
  • Avoid computing unused columns in subqueries.
  • Improvements to the push-down optimization.
• Add the ability for application-defined SQL functions to have the same name as join keywords: CROSS, FULL, INNER, LEFT, NATURAL, OUTER, or RIGHT.
• Enhancements to PRAGMA integrity_check:
  • Detect and raise an error when a NaN value is stored in a NOT NULL column.
  • Improved error message output identifies the root page of a b-tree when an error is found within a b-tree.
• Allow the session extension to be configured to capture changes from tables that lack an explicit ROWID.
• Added the subsecond modifier to the date and time functions.
• Negative values passed into sqlite3_sleep are henceforth interpreted as 0.
• The maximum recursion depth for JSON arrays and objects is lowered from 2000 to 1000.
• Extended the built-in printf() function so the comma option now works with floating-point conversions in addition to integer conversions.
• Miscellaneous bug fixes and performance optimizations

• Multiple fixes for reads past the end of memory buffers (NB: reads not writes) in the following circumstances:
  • When processing a corrupt database file.
  • In the recovery extension.
  • In FTS3 when processing a corrupt database file.
• Fix sqlite3_error_offset so that it does not return out-of-range values when reporting errors associated with generated columns.
• Multiple fixes in the query optimizer for problems that cause incorrect results for bizarre, fuzzer-generated queries.
• Increase the size of the reference counter in the page cache object to 64 bits to ensure that the counter never overflows.
• Fix a performance regression caused by a bug fix in patch release v5.43.1.

• Ensure that the datatype for column t1.x in CREATE TABLE t1 AS SELECT CAST(7 AS INT) AS x; continues to be INT and is not NUM, for historical compatibility.
• Enhance PRAGMA integrity_check to detect when extra bytes appear at the end of an index record.
• Fix various obscure bugs.

• Query planner improvements:
  • Make use of indexed expressions within an aggregate query that includes a GROUP BY clause.
  • The query planner has improved awareness of when an index is a covering index and adjusts predicted runtimes accordingly.
  • The query planner is more aggressive about using co-routines rather than materializing subqueries and views.
  • Queries against the built-in table-valued functions json_tree and json_each will now usually treat “ORDER BY rowid” as a no-op.
  • Enhance the ability of the query planner to use indexed expressions even if the expression has been modified by the constant-propagation optimization.
• Add the built-in unhex SQL function.
• Add the base64 and base85 application-defined functions as an extension.
• Add the sqlite3_stmt_scanstatus_v2 interface.
• In-memory databases created using sqlite3_deserialize now report their filename as an empty string, not as 'x'.
• Enhance the PRAGMA integrity_check command so that it detects when text strings in a table are equivalent to but not byte-for-byte identical to the same strings in the index.
• Enhance the carray table-valued function so that it is able to bind an array of BLOB objects.
• Added the sqlite3_is_interrupted interface.
• Long-running calls to sqlite3_prepare and similar now invoke the progress handler callback and react to sqlite3_interrupt.
• The sqlite3_vtab_in_first and sqlite3_vtab_in_next functions are enhanced so that they reliably detect if they are invoked on a parameter that was not selected for multi-value IN processing using sqlite3_vtab_in. They return SQLITE_ERROR instead of SQLITE_MISUSE in this case.
• The parser now ignores excess parentheses around a subquery on the right-hand side of an IN operator, so that SQLite now works the same as PostgreSQL in this regard. Formerly, SQLite treated the subquery as an expression with an implied “LIMIT 1”.
• Added the SQLITE_FCNTL_RESET_CACHE option to the sqlite3_file_control API.
• Fix problems with large compressed LSM databases.
• Miscellaneous performance enhancements.
• Remove the long-deprecated FTS1 and FTS2 full text search modules.

• Fix a potential infinite loop in the memsys5 alternative memory allocator.
• Various other obscure fixes.

• Add the recovery extension that might be able to recover some content from a corrupt database file.
• Query planner enhancements:
  • Recognize covering indexes on tables with more than 63 columns where columns beyond the 63rd column are used in the query and/or are referenced by the index.
  • Extract the values of expressions contained within expression indexes where practical, rather than recomputing the expression.
  • The NOT NULL and IS NULL operators (and their equivalents) avoid loading the content of large strings and BLOB values from disk.
  • Avoid materializing a view on which a full scan is performed exactly once. Use and discard the rows of the view as they are computed.
  • Allow flattening of a subquery that is the right-hand operand of a LEFT JOIN in an aggregate query.
• A new typedef named sqlite3_filename is added and used to represent the name of a database file. Various interfaces are modified to use the new typedef instead of C_char_ptr. This interface change should be fully backwards compatible, though it might cause (harmless) compiler warnings when rebuilding some legacy applications.
• Add the sqlite3_value_encoding interface.
• Security enhancement: SQLITE_DBCONFIG_DEFENSIVE is augmented to prohibit changing the schema_version. The schema_version becomes read-only in defensive mode.
• Enhancements to the PRAGMA integrity_check statement:
  • Columns in non-STRICT tables with TEXT affinity should not contain numeric values.
  • Columns in non-STRICT tables with NUMERIC affinity should not contain TEXT values that could be converted into numbers.
  • Verify that the rows of a WITHOUT ROWID table are in the correct order.
• Enhance the VACUUM INTO statement so that it honors the PRAGMA synchronous setting.
• Enhance the sqlite3_strglob and sqlite3_strlike APIs so that they are able to accept nil pointers for their string parameters and still generate a sensible result.
• Change the algorithm used by SQLite's built-in pseudo-random number generator (PRNG) from RC4 to Chacha20.
• Allow two or more indexes to have the same name as long as they are all in separate schemas.
• Miscellaneous performance optimizations result in about 1% fewer CPU cycles used on typical workloads.

This is a security release that addresses a single long-standing problem in the FTS3 extension. An attacker who is able to execute arbitrary SQL statements and who can create a corrupt database that is 2GB or larger in size might be able to trick FTS3 into overflowing an integer used to size a memory allocation, causing the allocation to be too small and ultimately resulting in a buffer overrun. It also includes fixes for other obscure weaknesses.

• Fix a long-standing problem in the btree balancer that might, in rare cases, cause database corruption if the application uses an application-defined page cache.
• Enhance SQLITE_DBCONFIG_DEFENSIVE so that it disallows CREATE TRIGGER statements if one or more of the statements in the body of the trigger write into shadow tables.
• Fix a possible integer overflow in the size computation for a memory allocation in FTS3.

• Use a statement journal on DML statement affecting two or more database rows if the statement makes use of SQL functions that might abort.
• Use a mutex to protect the PRAGMA temp_store_directory and PRAGMA data_store_directory statements, even though they are deprecated and documented as not being threadsafe.
• Other bug and warning fixes.

• Fix a performance regression in the query planner associated with rearranging the order of FROM clause terms in the presences of a LEFT JOIN.
• Apply fixes for CVE-2022-35737 and other minor problems discovered by internal testing.

• Fix an incorrect result from a query that uses a view that contains a compound SELECT in which only one arm contains a RIGHT JOIN and where the view is not the first FROM clause term of the query that contains the view.
• Fix a long-standing problem with ALTER TABLE RENAME that can only arise if the sqlite3_limit(SQLITE_LIMIT_SQL_LENGTH) is set to a very small value.
• Enhance the sqlite_stmt virtual table so that it buffers all of its output.

• Add support for RIGHT and FULL OUTER JOIN.
• Add new binary comparison operators IS NOT DISTINCT FROM and IS DISTINCT FROM that are equivalent to IS and IS NOT, respective, for compatibility with PostgreSQL and SQL standards.
• Add a new return code (value “3”) from the sqlite3_vtab_distinct interface that indicates a query that has both DISTINCT and ORDER BY clauses.
• Added the sqlite3_db_name interface.
• Defer materializing views until the materialization is actually needed, thus avoiding unnecessary work if the materialization turns out to never be used.
• The HAVING clause of a SELECT statement is now allowed on any aggregate query, even queries that do not have a GROUP BY clause.
• Many microoptimizations collectively reduce CPU cycles by about 2.3%.

• Fix a problem with the new Bloom filter optimization that might cause an incorrect answer when doing a LEFT JOIN with a WHERE clause constraint that says that one of the columns on the right table of the LEFT JOIN is NULL.
• Other minor patches.

• Fix problems with the new Bloom filter optimization that might cause some obscure queries to get an incorrect answer.
• Fix the localtime modifier of the date and time functions so that it preserves fractional seconds.
• Fix the sqlite_offset SQL function so that it works correctly even in corner cases such as when the argument is a virtual column or the column of a view.
• Fix row value IN operator constraints on virtual tables so that they work correctly even if the virtual table implementation relies on bytecode to filter rows that do not satisfy the constraint.

• Added the LSM database and the LSM virtual table extension.
• JSON functions are now built-ins; JSON is on by default. It is no longer necessary to call sqlite3_json_init(), that function is now deprecated and does nothing.
• Added the -> and ->> operators for easier processing of JSON. The new operators are compatible with MySQL and PostgreSQL.
• Enhancements to date and time functions:
  • Added the unixepoch() function.
  • Added the auto modifier and the julianday modifier.
• Rename the printf() SQL function to format for better compatibility. The original printf() name is retained as an alias for backwards compatibility.
• Added the sqlite3_error_offset interface, which can sometimes help to localize an SQL error to a specific character in the input SQL text, so that applications can provide better error messages.
• Enhanced the interface to virtual tables as follows:
  • Added the sqlite3_vtab_distinct interface.
  • Added the sqlite3_vtab_rhs_value interface.
  • Added new operator types SQLITE_INDEX_CONSTRAINT_LIMIT and SQLITE_INDEX_CONSTRAINT_OFFSET.
  • Added the sqlite3_vtab_in interface (and related) to enable a virtual table to process IN operator constraints all at once, rather than processing each value of the right-hand side of the IN operator separately.
• Query planner enhancements:
  • Use a Bloom filter to speed up large analytic queries.
  • Use a balanced merge tree to evaluate UNION or UNION ALL compound SELECT statements that have an ORDER BY clause.
• The ALTER TABLE statement is changed to silently ignores entries in the sqlite_schema table that do not parse when PRAGMA writable_schema=ON.

• Fix a bug that can cause database corruption if a SAVEPOINT is rolled back while in PRAGMA temp_store=MEMORY mode, and other changes are made, and then the outer transaction commits.
• Fix a long-standing problem with ON DELETE CASCADE and ON UPDATE CASCADE in which a cache of the bytecode used to implement the cascading change was not being reset following a local DDL change.

• Fix a bug in UPSERT that can cause incorrect byte-code to be generated for some obscure but valid SQL, possibly resulting in a nil-pointer dereference.
• Fix an out of bounds read that can occur in FTS5 when reading corrupt database files.

• STRICT tables provide a prescriptive style of data type management, for developers who prefer that kind of thing.
• When adding columns that contain a CHECK constraint or a generated column containing a NOT NULL constraint, the ALTER TABLE ADD COLUMN now checks new constraints against preexisting rows in the database and will only proceed if no constraints are violated.
• Added the PRAGMA table_list statement.
• Added the sqlite3_autovacuum_pages interface.
• The sqlite3_deserialize does not and has never worked for the TEMP database. That limitation is now noted in the documentation.
• The query planner now omits ORDER BY clauses on subqueries and views if removing those clauses does not change the semantics of the query.
• The generate_series table-valued function extension is modified so that the first parameter (“START”) is now required. This is done as a way to demonstrate how to write table-valued functions with required parameters.
• Added new sqlite3_changes64 and sqlite3_total_changes64 interfaces.
• Added the SQLITE_OPEN_EXRESCODE flag option to sqlite3_open_v2.
• Use less memory to hold the database schema.

• Support Delphi 11 Alexandria Win32 and Win64.

Incremental feature and performance improvements:

• Improvement to the EXPLAIN QUERY PLAN output to make it easier to understand.
• Byte-order marks at the start of a token are skipped as if they were whitespace.
• An error is raised on any attempt to access the rowid of a VIEW or subquery. Formerly, the rowid of a VIEW would be indeterminate and often would be NULL.
• The “memdb” VFS now allows the same in-memory database to be shared among multiple database connections in the same process as long as the database name begins with “/”.
• Back out the EXISTS-to-IN optimization as it was found to slow down queries more often than speed them up.
• Improve the constant-propagation optimization so that it works on non-join queries.

New APIs:

• sqlite3_preupdate_blobwrite: Return the index of the column being written in a preupdate callback.
• sqlite3session_object_config: Configure a session object.
• sqlite3session_changeset_size: Return an upper-limit for the size of a session changeset.

• Fix defects in the new ALTER TABLE DROP COLUMN feature that could corrupt the database file.
• Fix an obscure query optimizer problem that might cause an incorrect query result.

Yet another patch release to fix obscure problems in features associated with SQLite 3.35.0:

• Fix a segmentation fault if the new RETURNING syntax contains “table.*”.
• Fix a defect in the query planner optimization attempting to process EXISTS operators in the WHERE clause as if they were IN operators.
• Fix the new RETURNING feature so that it raises an error if one of the terms in the RETURNING clause references a unknown table, instead of silently ignoring that error.

Fix a handful of problems discovered in prior releases:

• Enhance the OP_OpenDup opcode of the bytecode engine so that it works even if the cursor being duplicated itself came from OP_OpenDup. This problem only came to light due to the recent MATERIALIZED hint enhancement.
• When materializing correlated common table expressions, do so separately for each use case, as that is required for correctness. This fixes a problem that was introduced by the MATERIALIZED hint enhancement.
• Improvements to error messages generated by faulty common table expressions.
• Fix to the SELECT statement syntax diagram so that the FROM clause syntax is shown correctly.
• Improvements the xBestIndex method in the implementation of the wholenumber virtual table extension so that it does a better job of convincing the query planner to avoid trying to materialize a table with an infinite number of rows.

This is a patch release to fix problems that were discovered shortly after the previous release:

• Fix a problem with growing databases in the Append VFS extension.
• Ensure that date/time functions with no arguments (which generate responses that depend on the current time) are treated as non-deterministic functions.
• Limit the new UNION ALL optimization so that it does not try to make too many new subqueries.

• Fix a bug in the new ALTER TABLE DROP COLUMN feature when used on columns that are indexed and that are quoted in the index definition.

• Added built-in SQL math functions.
• Added support for ALTER TABLE DROP COLUMN.
• Generalize UPSERT:
  • Allow multiple ON CONFLICT clauses that are evaluated in order,
  • The final ON CONFLICT clause may omit the conflict target and yet still use DO UPDATE.
• Add support for the RETURNING clause on DELETE, INSERT, and UPDATE statements.
• Use less memory when running VACUUM on databases containing very large TEXT or BLOB values. It is no longer necessary to hold the entire TEXT or BLOB in memory all at once.
• Add support for the MATERIALIZED and NOT MATERIALIZED hints when specifying common table expressions. The default behavior was formerly NOT MATERIALIZED, but is now changed to MATERIALIZED for CTEs that are used more than once.
• The SQLITE_DBCONFIG_ENABLE_TRIGGER and SQLITE_DBCONFIG_ENABLE_VIEW settings are modified so that they only control triggers and views in the main database schema or in attached database schemas and not in the TEMP schema. TEMP triggers and views are always allowed.
• Query planner/optimizer improvements:
  • Enhancements to the min/max optimization so that it works better with the IN operator and the OP_SeekScan optimization of the previous release.
  • Attempt to process EXISTS operators in the WHERE clause as if they were IN operators, in cases where this is a valid transformation and seems likely to improve performance.
  • Allow UNION ALL sub-queries to be flattened even if the parent query is a join.
  • Use an index, if appropriate, on IS NOT NULL expressions in the WHERE clause, even if STAT4 is disabled.
  • Expressions of the form “x IS NULL” or “x IS NOT NULL” might be converted to simply FALSE or TRUE, if “x” is a column that has a “NOT NULL” constraint and is not involved in an outer join.
  • Avoid checking foreign key constraints on an UPDATE statement if the UPDATE does not modify any columns associated with the foreign key.
  • Allow WHERE terms to be pushed down into sub-queries that contain window functions, as long as the WHERE term is made up of entirely of constants and copies of expressions found in the PARTITION BY clauses of all window functions in the sub-query.

Bug fixes:

• Fix a bug in the IN-operator optimization that can cause an incorrect answer.
• Fix incorrect answers from the LIKE operator if the pattern ends with “%” and there is an “ESCAPE '_'” clause.
• Fix a long-standing bug with Delphi Win64 which prevented SQL parameter extension in trace and sqlite3_expanded_sql.

• Fix a potential use-after-free bug when processing a a subquery with both a correlated WHERE clause and a “HAVING 0” clause and where the parent query is an aggregate.
• Fix minor problems in extensions.

• New TDISQLite3UniDirQuery.IntegerPrimaryKeyField option to work around the 32-bit limitation of Delphi's ftAutoInc DataType and its corresponding TAutoIncField. Data loss may occur if an INTEGER PRIMARY KEY column uses the full 64-bit range available to SQLite3 integers.
• Added the sqlite3_txn_state interface for reporting on the current transaction state of the database connection.
• Enhance recursive common table expressions to support two or more recursive terms as is done by SQL Server, since this helps make queries against graphs easier to write and faster to execute.
• Improved error messages on CHECK constraint failures.
• Query planner improvements:
  • Improved estimates for the cost of running a DISTINCT operator.
  • When doing an UPDATE or DELETE using a multi-column index where only a few of the earlier columns of the index are useful for the index lookup, postpone doing the main table seek until after all WHERE clause constraints have been evaluated, in case those constraints can be covered by unused later terms of the index, thus avoiding unnecessary main table seeks.
  • The new OP_SeekScan opcode is used to improve performance of multi-column index look-ups when later columns are constrained by an IN operator.
• The BEGIN IMMEDIATE and BEGIN EXCLUSIVE commands now work even if one or more attached database files are read-only.
• Enhanced FTS5 to support trigram indexes.
• Improved performance of WAL mode locking primitives in cases where there are hundreds of connections all accessing the same database file at once.
• Enhanced the carray() table-valued function to include a single-argument form that is bound using the auxiliary sqlite3_carray_bind interface.
• The substr() SQL function can now also be called “substring()” for compatibility with SQL Server.
• The syntax diagrams are now rendered as SVG for improved legibility and ease of maintenance. SVG requires at least Internet Explorer 9. If you experience display problems, you may need to update.

• Support for UPDATE FROM following the PostgreSQL syntax.
• Increase the maximum size of database files to 281 TB.
• Extended the PRAGMA integrity_check statement so that it can optionally be limited to verifying just a single table and its indexes, rather than the entire database file.
• Added the decimal extension for doing arbitrary-precision decimal arithmetic.
• Enhancements to the ieee754 extension for working with IEEE 754 binary64 numbers.
• Query planner improvements:
  1. Add the ability to find a full-index-scan query plan for queries using INDEXED BY which previously would fail with “no query solution”.
  2. Do a better job of detecting missing, incomplete, and/or dodgy sqlite_stat1 data and generates good query plans in spite of the misinformation.
  3. Improved performance of queries like “SELECT min(x) FROM t WHERE y IN (?,?,?)” assuming an index on t(x,y).
• In WAL mode, if a writer crashes and leaves the shm file in an inconsistent state, subsequent transactions are now able to recover the shm file even if there are active read transactions. Before this enhancement, shm file recovery that scenario would result in an SQLITE_PROTOCOL error.

• Various minor bug fixes.

• Fix a long-standing bug in the byte-code engine that can cause a COMMIT command report as success when in fact it failed to commit. The problem was present since SQLite 3.17.0 (2017-02-13). If you were to retry a COMMIT command over and over after it returned SQLITE_BUSY, it might have eventually reported success, even though it was still blocked.
• Updates to demo projects.

• Support Delphi 10.4 Sydney Win32 and Win64.

• Added support for approximate ANALYZE using the PRAGMA analysis_limit command.
• Added the bytecode virtual table.
• Add the checksum VFS shim to the set of run-time loadable extensions included in the source tree.
• Added the iif() SQL function.
• INSERT and UPDATE statements now always apply column affinity before computing CHECK constraints. This bug fix could, in theory, cause problems for legacy databases with unorthodox CHECK constraints the require the input type for an INSERT is different from the declared column type.
• Added the sqlite3_create_filename, sqlite3_free_filename, and sqlite3_database_file_object interfaces to better support of VFS shim implementations.
• Increase the default upper bound on the number of parameters from 999 to 32766.
• Added code for the UINT collating sequence as an optional loadable extension.
• The ESCAPE clause of a LIKE operator now overrides wildcard characters, so that the behavior now matches what PostgreSQL does.
• Fix two long-standing bugs that allow malicious SQL statements to crash the process that is running SQLite.

• DISQLite3's UTF-16 to UTF-8 conversions (and vice versa) now convert Unicode Code Points from $100000 to $10FFFF (Supplemental Planes). Previously, conversion was limited to the Basic Multilingual Plane (BMP) from $0000 to $FFFF. This primarily improves sqlite3_decode_utf8 and sqlite3_encode_utf8, but also internal conversions, i.e. in TDISQLite3UniDirQuery, which use these functions.
• Follow up renaming parameter names which caused a compile error in Delphi 2007 and before.

• Add support for generated columns.
• Add the sqlite3_hard_heap_limit64 interface and the corresponding PRAGMA hard_heap_limit command.
• Enhance the function_list pragma to show the number of arguments on each function, the type of function (scalar, aggregate, window), and the function property flags SQLITE_DETERMINISTIC, SQLITE_DIRECTONLY, SQLITE_INNOCUOUS, and/or SQLITE_SUBTYPE.
• Add the aggregated mode feature to the DBSTAT virtual table.
• Add the SQLITE_OPEN_NOFOLLOW option to sqlite3_open_v2 that prevents SQLite from opening symbolic links.
• Added the “#-N” array notation for JSON function path arguments.
• Added the SQLITE_DBCONFIG_TRUSTED_SCHEMA connection setting which is also controllable via the new trusted_schema pragma.
• Added APIs sqlite3_filename_database, sqlite3_filename_journal, and sqlite3_filename_wal which are useful for specialized extensions.
• Add the sqlite3_uri_key interface.
• Upgraded the sqlite3_uri_parameter function so that it works with the rollback journal or WAL filename in addition to the database filename.
• Provide the ability to tag application-defined SQL functions with new properties SQLITE_INNOCUOUS or SQLITE_DIRECTONLY.
• Add new verbs to sqlite3_vtab_config so that the xConnect method of virtual tables can declare the virtual table as SQLITE_VTAB_INNOCUOUS or SQLITE_VTAB_DIRECTONLY.
• Faster response to sqlite3_interrupt.
• Added the UUID extension module implementing functions for processing RFC-4122 UUIDs.
• Added the urifuncs SQL function extension.
• The lookaside memory allocator is enhanced to support two separate memory pools with different sized allocations in each pool. This allows more memory allocations to be covered by lookaside while at the same time reducing the heap memory usage to 48KB per connection, down from 120KB.
• The legacy_file_format pragma is deactivated. It is now a no-op. In its place, the SQLITE_DBCONFIG_LEGACY_FILE_FORMAT option to sqlite3_db_config is provided. The legacy_file_format pragma is deactivated because (1) it is rarely useful and (2) it is incompatible with VACUUM in schemas that have tables with both generated columns and descending indexes.
• TDISQLite3UniDirQuery: ftMemo and ftWideMemo fields return string representations of SQLITE_INTEGER and SQLITE_FLOAT content.
• For unknown reasons, the following Delphi compilers crash when creating .hpp and .obj with the -JPHNE option:
  • Win32: Delphi 2005. No error message.
  • Win64: Delphi XE3, Delphi XE4, Delphi XE4, Delphi XE5, Delphi XE6, Delphi XE7, Delphi XE8, Delphi 10.0 Seattle, Delphi 10.1 Berlin. Error message: “../../../shared/util-common/elfout.c:300: Assertion `(ef->sht[shndx].sh_type == SHT_NOBITS) && (ftell (ef->fp) == (long)ef->sht[shndx].sh_offset)' failed.” followed by “Abnormal program termination”.
  • .hpp and .obj files are not included for these compilers.

• Add support for the FILTER clause on aggregate functions.
• Add support for the NULLS FIRST and NULLS LAST syntax in ORDER BY clauses.
• The index_info and index_xinfo pragmas are enhanced to provide information about the on-disk representation of WITHOUT ROWID tables.
• Add the sqlite3_drop_modules interface, allowing applications to disable automatically loaded virtual tables that they do not need.
• Enhance the RBU extension to support indexes on expressions.
• Change the schema parser so that it will error out if any of the type, name, and tbl_name columns of the sqlite_schema table have been corrupted and the database connection is not in writable_schema mode.
• Add the SQLITE_DBCONFIG_ENABLE_VIEW option for sqlite3_db_config.
• Added the SQLITE_DIRECTONLY flag for application-defined SQL functions to prevent those functions from being used inside triggers and views.

• Added the SQLITE_DBCONFIG_DQS_DML and SQLITE_DBCONFIG_DQS_DDL actions to sqlite3_db_config for activating and deactivating the double-quoted string literal misfeature. Both default to “on” for legacy compatibility, but developers are encouraged to turn them “off”.
• Improvements to the query planner:
  • Improved optimization of AND and OR operators when one or the other operand is a constant.
  • Enhancements to the LIKE optimization for cases when the left-hand side column has numeric affinity.
• Added the sqlite_dbdata virtual table for extracting raw low-level content from an SQLite database, even a database that is corrupt.

• Enhanced window functions:
  • Add support for the EXCLUDE clause.
  • Add support for window chaining.
  • Add support for GROUPS frames.
  • Add support for “<expr> PRECEDING” and “<expr> FOLLOWING” boundaries in RANGE frames.
• Added the new sqlite3_stmt_isexplain interface for determining whether or not a prepared statement is an EXPLAIN.
• Enhanced VACUUM INTO so that it works for read-only databases.
• New query optimizations:
  • Enable the LIKE optimization for cases when the ESCAPE keyword is present and PRAGMA case_sensitive_like is on.
  • In queries that are driven by a partial index, avoid unnecessary tests of the constraint named in the WHERE clause of the partial index, since we know that constraint must always be true.
• Added the fossildelta extension that can create, apply, and deconstruct the Fossil DVCS file delta format that is used by the RBU extension.
• Added the SQLITE_DBCONFIG_WRITABLE_SCHEMA verb for the sqlite3_db_config interface, that does the same work as PRAGMA writable_schema without using the SQL parser.
• Added the sqlite3_value_frombind API for determining if the argument to an SQL function is from a bound parameter.
• Added the blobio extension.
• Security and compatibilities enhancements to fts3_tokenizer():
  • The fts3_tokenizer() function always returns NULL unless either the legacy application-defined FTS3 tokenizers interface are enabled using the sqlite3_db_config(SQLITE_DBCONFIG_ENABLE_FTS3_TOKENIZER) setting, or unless the first argument to fts3_tokenizer() is a bound parameter.
  • The two-argument version of fts3_tokenizer() accepts a pointer to the tokenizer method object even without the sqlite3_db_config(SQLITE_DBCONFIG_ENABLE_FTS3_TOKENIZER) setting if the second argument is a bound parameter
• Improved robustness against corrupt database files.
• Miscellaneous performance enhancements

• Added the VACUUM INTO command.
• Added the prefixes extension.
• Issue an SQLITE_WARNING message on the error log if a double-quoted string literal is used.
• The sqlite3_normalized_sql interface works on any prepared statementcreated using sqlite3_prepare_v2 or sqlite3_prepare_v3. It is no longer necessary to use sqlite3_prepare_v3 with SQLITE_PREPARE_NORMALIZEin order to use sqlite3_normalized_sql.
• Added the remove_diacritics=2 option to FTS3 and FTS5.
• Added the SQLITE_PREPARE_NO_VTAB option to sqlite3_prepare_v3. Use that option to prevent circular references to shadow tables from causing resource leaks.
• Enhancements to the sqlite3_deserialize interface:
  • Add the SQLITE_FCNTL_SIZE_LIMIT file-control for setting an upper bound on the size of the in-memory database created by sqlite3_deserialize. The default upper bound is 1GiB, or whatever alternative value is specified by sqlite3_config(SQLITE_CONFIG_MEMDB_MAXSIZE).
  • Honor the SQLITE_DESERIALIZE_READONLY flag, which was previously described in the documentation, but was previously a no-op.
• Increased robustness against malicious SQL that is run against a maliciously corrupted database.

Bug fixes:

• Do not use a partial index to do a table scan on an IN operator.
• Fix the query flattener so that it works on queries that contain subqueries that use window functions.
• Ensure that ALTER TABLE modifies table and column names embedded in WITH clauses that are part of views and triggers.
• Fix a parser bug that prevented the use of parentheses around table-valued functions.
• Fix a problem with the OR optimization on indexes on expressions.
• Fix a problem with the LEFT JOIN strength reduction optimization in which the optimization was being applied inappropriately due to an IS NOT NULL operator.
• Fix the REPLACE command so that it is no longer able to sneak a NULL value into a NOT NULL column even if the NOT NULL column has a default value of NULL.
• Fix a problem with the use of window functions used within correlated subqueries.
• Fix the ALTER TABLE RENAME COLUMN command so that it works for tables that have redundant UNIQUE constraints.
• Fix a bug that caused zeroblob values to be truncated when inserted into a table that uses an expression index.
• Fix a bug in the query optimizer: an adverse interaction between the OR optimization and the optimization that tries to use values read directly from an expression index instead of recomputing the expression.
• Fix a bug causing a crash when a window function is misused.

• Fix sqlite3_rekey and sqlite3_rekey_v2 encryption errors with databases greater than about 4 MB.

• Support Delphi 10.3 Rio Win32 and Win64.

Enhancements:

• Add support for window functions.
• Enhancements the ALTER TABLE command:
  • Add support for renaming columns within a table using ALTER TABLE table RENAME COLUMN oldname TO newname.
• Add support for PostgreSQL-style UPSERT.
  • Fix table rename feature so that it also updates references to the renamed table in triggers and views.
• Query optimizer improvements:
  • Avoid unnecessary loads of columns in an aggregate query that are not within an aggregate function and that are not part of the GROUP BY clause.
  • The IN-early-out optimization: When doing a look-up on a multi-column index and an IN operator is used on a column other than the left-most column, then if no rows match against the first IN value, check to make sure there exist rows that match the columns to the right before continuing with the next IN value.
  • Use the transitive property to try to propagate constant values within the WHERE clause. For example, convert “a=99 AND b=a” into “a=99 AND b=99”.
• Enhance the PRAGMA integrity_check command for improved detection of problems on the page freelist.
• Allow the xBestIndex method of virtual table implementations to return SQLITE_CONSTRAINT to indicate that the proposed query plan is unusable and should not be given further consideration.
• Added the SQLITE_DBCONFIG_DEFENSIVE option which disables the ability to create corrupt database files using ordinary SQL.
• Added support for read-only shadow tables when the SQLITE_DBCONFIG_DEFENSIVE option is enabled.
• Added the PRAGMA legacy_alter_table command, which if enabled causes the ALTER TABLE command to behave like older version of SQLite (prior to version 3.25.0) for compatibility.
• Added PRAGMA table_xinfo that works just like PRAGMA table_info except that it also shows hidden columns in virtual tables.
• Added the EXPLAIN virtual table as a run-time loadable extension.
• Add a limit counter to the query planner to prevent excessive sqlite3_prepare times for certain pathological SQL inputs.
• Added support for the sqlite3_normalized_sql interface.
• Enhanced triggers so that they can use table-valued functions that exist in schemas other than the schema where the trigger is defined.
• Added the SQLITE_FCNTL_DATA_VERSION file-control.
• Added the Geopoly module.
• Add support for auxiliary columns in r-tree tables.
• Add APIs for discovering SQL keywords used by SQLite: sqlite3_keyword_count, sqlite3_keyword_name, and sqlite3_keyword_check.
• Add APIs for dynamic strings based on the sqlite3_str object.
• Enhance ALTER TABLE so that it recognizes “true” and “false” as valid arguments to DEFAULT.
• Add the sorter-reference optimization as a compile-time option.
• Improve the format of the EXPLAIN QUERY PLAN raw output, so that it gives better information about the query plan and about the relationships between the various components of the plan.
• Added the SQLITE_DBCONFIG_RESET_DATABASE option to the sqlite3_db_config API.
• Enhancements to the session extension:
  • Added the SQLITE_CHANGESETAPPLY_INVERT flag
  • Added the sqlite3changeset_start_v2 interface and the SQLITE_CHANGESETSTART_INVERT flag.

Performance:

• Optimization: When doing an UPDATE on a table with indexes on expressions, do not update the expression indexes if they do not refer to any of the columns of the table being updated.
• UPDATE avoids unnecessary low-level disk writes when the contents of the database file do not actually change. For example, “UPDATE t1 SET x=25 WHERE y=?” generates no extra disk I/O if the value in column x is already 25. Similarly, when doing UPDATE on records that span multiple pages, only the subset of pages that actually change are written to disk. This is a low-level performance optimization only and does not affect the behavior of TRIGGERs or other higher level SQL structures.
• Queries that use ORDER BY and LIMIT now try to avoid computing rows that cannot possibly come in under the LIMIT. This can greatly improve performance of ORDER BY LIMIT queries, especially when the LIMIT is small relative to the number of unrestricted output rows.
• The OR optimization is allowed to proceed even if the OR expression has also been converted into an IN expression. Uses of the OR optimization are now also more clearly shown in the EXPLAIN QUERY PLAN output.
• The query planner is more aggressive about using automatic indexes for views and subqueries for which it is not possible to create a persistent index.
• Make use of the one-pass UPDATE and DELETE query plans in the R-Tree extension where appropriate.
• Performance improvements in the parser.

Bug fixes:

• The ORDER BY LIMIT optimization might have caused an infinite loop in the byte code of the prepared statement under very obscure circumstances, due to a confluence of minor defects in the query optimizer.
• On an UPSERT when the order of constraint checks is rearranged, ensure that the affinity transformations on the inserted content occur before any of the constraint checks.
• The LIKE optimization was generating incorrect byte-code and hence getting the wrong answer if the left-hand operand has numeric affinity and the right-hand-side pattern is '/%' or if the pattern begins with the ESCAPE character.
• For the right-hand table of a LEFT JOIN, compute the values of expressions directly rather than loading precomputed values out of an expression index as the expression index might not contain the correct value.
• Do not attempt to use terms from the WHERE clause to enable indexed lookup of the right-hand table of a LEFT JOIN.
• Fix a memory leak that can occur following a failure to open error in the CSV virtual table.
• Fix a long-standing problem wherein a corrupt schema on the sqlite_sequence table used by AUTOINCREMENT can lead to a crash.
• Fix the json_each() function so that it returns valid results on its “fullkey” column when the input is a simple value rather than an array or object.
• Fix a memory leak when reading TDISQLite3UniDirQuery string fields.

• Performance enhancements in the parser.
• Add the sqlite3_serialize and sqlite3_deserialize interfaces.
• Recognize TRUE and FALSE as constants. (For compatibility, if there exist columns named “true” or “false”, then the identifiers refer to the columns rather than Boolean constants.)
• Support operators IS TRUE, IS FALSE, IS NOT TRUE, and IS NOT FALSE.
• Added the SQLITE_DBSTATUS_CACHE_SPILL option to sqlite3_db_status for reporting the number of cache spills that have occurred.
• The “alternate-form-2” flag (“!”) on the built-in printf implementation now causes string substitutions to measure the width and precision in characters instead of bytes.
• If the xColumn method in a virtual table implementation returns an error message using sqlite3_result_error then give that error message preference over internally-generated messages.
• Add support for INSERT OR REPLACE, INSERT OR IGNORE, and UPDATE OR REPLACE in the Zipfile virtual table.
• Enhance the sqlite3changeset_apply interface so that it is hardened against attacks from deliberately corrupted changeset objects.
• Added the sqlite3_normalize extension function.
• Query optimizer enhancements:
  • Improve the omit-left-join optimization so that it works in cases where the right-hand table is UNIQUE but not necessarily NOT NULL.
  • Improve the push-down optimization so that it works for many LEFT JOINs.
  • Avoid unnecessary writes to the sqlite_sequence table when an AUTOINCREMENT table is updated with an rowid that is less than the maximum.
  • Add the LEFT JOIN strength reduction optimization that converts a LEFT JOIN into an ordinary JOIN if there exist terms in the WHERE clause that would prevent the extra all-NULL row of the LEFT JOIN from appearing in the output set.
• Bug fixes:
  • Fix two problems in the new LEFT JOIN strength reduction optimization.
  • Fix misbehavior of the FTS5 xBestIndex method.
  • Fix a harmless reference to an uninitialized virtual machine register.
  • Fix the eval() SQL function extension so that it works with PRAGMA empty_result_callbacks=ON.
  • Fix the generate_series virtual table so that it correctly returns no rows if any of its constraints are NULL.
  • Fix the parser to accept valid row value syntax.
  • Fix the query planner so that it takes into account dependencies in the arguments to table-valued functions in subexpressions in the WHERE clause.
  • Fix incorrect result with complex OR-connected WHERE and STAT4.
  • Fix potential corruption in indexes on expressions due to automatic datatype conversions.
  • Assertion fault in FTS4.
  • Incorrect result on the less-than operator in row values.
  • Always interpret non-zero floating-point values as TRUE, even if the integer part is zero.
  • Issue an error rather instead of an assertion-fault or null-pointer dereference when the sqlite_schema table is corrupted so that the sqlite_sequence table root page is really a btree-index page.
  • Fix the ANALYZE command so that it computes statistics on tables whose names begin with “sqlite”.
  • Fix a possible infinite loop on VACUUM for corrupt database files.
  • Disallow parameters in the WITH clause of triggers and views.
  • Fix a potential memory leak in row value processing.
  • Improve the performance of the replace() SQL function for cases where there are many substitutions on megabyte-sized strings.
  • Provide an appropriate error message when the sqlite_schema table contains a CREATE TABLE AS statement. Formerly this caused either an assertion fault or null pointer dereference.
  • Fix a problem with using the LIKE optimization on an INTEGER PRIMARY KEY.

• The output of sqlite3_trace_v2 now shows each individual SQL statement run within a trigger.
• Add the ability to read from WAL mode databases even if the application lacks write permission on the database and its containing directory, as long as the -shm and -wal files exist in that directory.
• Added the rtreecheck scalar SQL function to the R-Tree extension.
• Added the sqlite3_vtab_nochange and sqlite3_value_nochange interfaces to help virtual table implementations optimize UPDATE operations.
• Added the sqlite3_vtab_collation interface.
• Added support for the “^” initial token syntax in FTS5.
• New extensions:
  • The Zipfile virtual table can read and write a ZIP Archive.
  • The sqlite_btreeinfo eponymous virtual table for introspecting and estimating the sizes of the btrees in a database.
  • The Append VFS is a VFS shim that allows an SQLite database to be appended to some other file. This allows (for example) a database to be appended to an executable that then opens and reads the database.
  • The incremental_index_check is a read-only eponymous-only virtual table that checkes indexes.
  • SQLAR compatible SQL compress and decompress functions.
• Query planner enhancements:
  • The optimization that uses an index to quickly compute an aggregate min() or max() is extended to work with indexes on expressions.
  • The decision of whether to implement a FROM-clause subquery as a co-routine or using query flattening now considers whether the result set of the outer query is “complex” (if it contains functions or expression subqueries). A complex result set biases the decision toward the use of co-routines.
  • The planner avoids query plans that use indexes with unknown collating functions.
  • The planner omits unused LEFT JOINs even if they are not the right-most joins of a query.
• Other performance optimizations:
  • A smaller and faster implementation of text to floating-point conversion subroutine: sqlite3AtoF().
  • Faster SQL parser.
• Provide the sqlite_offset(X) SQL function that returns the byte offset into the database file to the beginning of the record holding value X.
• Bug fixes:
  • Infinite loop on an UPDATE that uses an OR operator in the WHERE clause.
  • Incorrect query results when the skip-ahead-distinct optimization is used.
  • Incorrect query results on a join with a ORDER BY DESC.
  • Inconsistent result set column names between CREATE TABLE AS and a simple SELECT.
  • Assertion fault when doing REPLACE on an index on an expression.
  • Assertion fault when doing an IN operator on a constant index.
  • Correct sqlite3_update_hook callback zDB argument type.

• Allow ATTACH and DETACH commands to work inside of a transaction.
• Allow WITHOUT ROWID virtual tables to be writable if the PRIMARY KEY contains exactly one column.
• Improved de-quoting of column names for CREATE TABLE AS statements with an aggregate query on the right-hand side.
• Enhanced the LIKE optimization so that it works with an ESCAPE clause.
• Enhanced PRAGMA integrity_check and PRAGMA quick_check to detect obscure row corruption that they were formerly missing. Also update both pragmas so that they return error text rather than SQLITE_CORRUPT when encountering corruption in records.
• The query planner now prefers to implement FROM-clause subqueries using co-routines rather using the query flattener optimization. Support for the use of co-routines for subqueries may no longer be disabled.
• Pass information about !=, IS, IS NOT, NOT NULL, and IS NULL constraints into the xBestIndex method of virtual tables.
• Enhanced the CSV virtual table so that it accepts the last row of input if the final new-line character is missing.
• Remove the rarely-used “scratch” memory allocator. Replace it with the SQLITE_CONFIG_SMALL_MALLOC configuration setting that gives SQLite a hint that large memory allocations should be avoided when possible.
• Added the swarm virtual table to the existing union virtual table extension.
• Added the sqlite_dbpage virtual table for providing direct access to pages of the database file.
• Add a new type of fts5vocab virtual table – “instance” – that provides direct access to an FTS5 full-text index at the lowest possible level.
• Miscellaneous microoptimizations reduce CPU usage by about 2.1%.

Bug fixes:

• Fix an obscure memory leak in sqlite3_result_pointer.
• Avoid a possible use-after-free error by deferring schema resets until after the query planner has finished running.
• Only use indexes-on-expressions to optimize ORDER BY or GROUP BY if the COLLATE is correct.
• Fix an assertion fault that was coming up when the expression in an index-on-expressions is really a constant.
• Fix an assertion fault that could occur following PRAGMA reverse_unordered_selects.
• Fix a segfault that can occur for queries that use table-valued functions in an IN or EXISTS subquery.
• Fix a potential integer overflow problem when compiling a particular horrendous common table expression.
• Fix a potential out-of-bound read when querying a corrupt database file.

• Update the text of error messages returned by sqlite3_errmsg for some error codes.
• Add new pointer passing interfaces.
• Backwards-incompatible changes to some extensions in order to take advantage of the improved security offered by the new pointer passing interfaces:
  • Extending FTS5 → requires sqlite3_bind_pointer to find the fts5_api pointer.
  • carray(PTR,N) → requires sqlite3_bind_pointer to set the PTR parameter.
  • remember(V,PTR) → requires sqlite3_bind_pointer to set the PTR parameter.
• Added the SQLITE_STMT virtual table extension.
• Added the COMPLETION extension – designed to suggest tab-completions for interactive user interfaces. This is a work in progress. Expect further enhancements in future releases.
• Added the UNION virtual table extension.
• The built-in date and time functions have been enhanced so that they can be used in CHECK constraints, in indexes on expressions, and in the WHERE clauses of partial indexes, provided that they do not use the 'now', 'localtime', or'utc' keywords. Futher information.
• Added the sqlite3_prepare_v3 and sqlite3_prepare16_v3 interfaces with the extra “prepFlags” parameters.
• Provide the SQLITE_PREPARE_PERSISTENT flag for sqlite3_prepare_v3 and use it to limit lookaside memory misuse by FTS3, FTS5, and the R-Tree extension.
• Added the PRAGMA secure_delete=FAST command. When secure_delete is set to FAST, old content is overwritten with zeros as long as that does not increase the amount of I/O. Deleted content might still persist on the free-page list but will be purged from all b-tree pages.
• Query planner enhancements:
  • When generating individual loops for each ORed term of an OR scan, move any constant WHERE expressions outside of the loop, as is done for top-level loops.
  • The query planner examines the values of bound parameters to help determine if a partial index is usable.
  • When deciding between two plans with the same estimated cost, bias the selection toward the one that does not use the sorter.
  • Evaluate WHERE clause constraints involving correlated subqueries last, in the hope that they never have be evaluated at all.
  • Do not use the flattening optimization for a sub-query on the RHS of a LEFT JOIN if that subquery reads data from a virtual table as doing so prevents the query planner from creating automatic indexes on the results of the sub-query, which can slow down the query.
• Add SQLITE_STMTSTATUS_REPREPARE, SQLITE_STMTSTATUS_RUN, and SQLITE_STMTSTATUS_MEMUSED options for the sqlite3_stmt_status interface.
• Provide PRAGMA functions for PRAGMA integrity_check, PRAGMA quick_check, and PRAGMA foreign_key_check.
• The SQLITE_DBCONFIG_ENABLE_QPSG run-time option enables the query planner stability guarantee.
• Miscellaneous optimizations result in a 2% reduction in CPU cycles used.

Bug Fixes:

• Fix the behavior of sqlite3_column_name for queries that use the flattening optimization so that the result is consistent with other queries that do not use that optimization, and with PostgreSQL, MySQL, and SQLServer.
• Fix the query planner so that it knows not to use automatic indexes on the right table of LEFT JOIN if the WHERE clause uses the IS operator.
• Ensure that the query planner knows that any column of a flattened LEFT JOIN can be NULL even if that column is labeled with “NOT NULL”.
• Fix rare false-positives in PRAGMA integrity_check when run on a database connection with ATTACH|attached databases.
• Fix a bug that causes an assertion fault if certain dodgy CREATE TABLE declarations are used.

• Emergency patch release to fix a bug in the auto_vacuum logic that can lead to database corruption. Though the bug is obscure and rarely encountered, upgrading is recommended for all users, and especially for users who turn on auto_vacuum.

• The SQLITE_READ authorizer callback is invoked once with a column name that is an empty string for every table referenced in a query from which no columns are extracted.
• When using an index on an expression, try to use expression values already available in the index, rather than loading the original columns and recomputing the expression.
• Enhance the flattening optimization so that it is able to flatten views on the right-hand side of a LEFT JOIN.
• Use replace() instead of char() for escaping newline and carriage-return characters embedded in strings in the dbdump extension.
• Avoid unnecessary foreign key processing in UPDATE statements that do not touch the columns that are constrained by the foreign keys.
• On a DISTINCT query that uses an index, try to skip ahead to the next distinct entry using the index rather than stepping through rows, when an appropriate index is available.
• Avoid unnecessary invalidation of sqlite3_blob handles when making changes to unrelated tables.
• Transfer any terms of the HAVING clause that use only columns mentioned in the GROUP BY clause over to the WHERE clause for faster processing.
• Reuse the same materialization of a VIEW if that VIEW appears more than once in the same query.
• Enhance PRAGMA integrity_check so that it identifies tables that have two or more rows with the same rowid.
• Enhance the FTS5 query syntax so that column filters may be applied to arbitrary expressions.
• Enhance the json_extract() function to cache and reuse parses of JSON input text.
• Added the anycollseq loadable extension that allows a generic SQLite database connection to read a schema that contains unknown and/or application-specific collating sequences.

Bug Fixes:

• Fix a problem in REPLACE that can result in a corrupt database containing two or more rows with the same rowid.
• Fix a problem in PRAGMA integrity_check that was causing a subsequent VACUUM to behave suboptimally.
• Fix the PRAGMA foreign_key_check command so that it works correctly with foreign keys on WITHOUT ROWID tables.
• Disallow leading zeros in numeric constants in JSON.
• Disallow control characters inside of strings in JSON.
• Limit the depth of recursion for JSON objects and arrays in order to avoid excess stack usage in the recursive descent parser.
• Fix bugs in the LEFT JOIN flattening optimization.

• Support Delphi 10.2 Tokyo Win32 and Win64.
• Added the PRAGMA optimize command.
• The SQLite version identifier returned by the sqlite_source_id() SQL function and the sqlite3_sourceid C API is now a 64-digit SHA3-256 hash instead of a 40-digit SHA1 hash.
• Added the json_patch() SQL function to the JSON1 extension.
• Enhance the LIKE optimization so that it works for arbitrary expressions on the left-hand side as long as the LIKE pattern on the right-hand side does not begin with a digit or minus sign.
• Added the sqlite3_set_last_insert_rowid interface and use the new interface in the FTS3, FTS4, and FTS5 extensions to ensure that the sqlite3_last_insert_rowid interface always returns reasonable values.
• Enhance PRAGMA integrity_check and PRAGMA quick_check so that they verify CHECK constraints.
• Enhance the query plans for joins to detect empty tables early and halt without doing unnecessary work. Enhance the sqlite3_mprintf family of interfaces and the printf SQL function to put comma separators at the thousands marks for integers, if the “,” format modifier is used in between the “%” and the “d” (example: “%,d”).
• Begin enforcing SQLITE_LIMIT_VDBE_OP. This can be used, for example, to prevent excessively large prepared statements in systems that accept SQL queries from untrusted users.
• Various performance improvements.
• Add the dbdump extension.

Bug Fixes:

• Ensure that indexed expressions with collating sequences are handled correctly.
• Fix a bug in the 'start of …' modifiers for the date and time functions.
• Fix a potential segfault in complex recursive triggers.
• In the RBU extension, add extra sync operations to avoid the possibility of corruption following a power failure.
• The sqlite3_trace_v2 output for nested SQL statements should always begin with a “–” comment marker.

• Better performance from the R-Tree extension.
  • Uses the sqlite3_blob key/value access object instead of SQL for pulling content out of R-Tree nodes.
  • Other miscellaneous enhancements such as loop unrolling.
• Increase the default lookaside size from 512,125 to 1200,100 as this provides better performance while only adding 56KB of extra memory per connection. Memory-sensitive applications can restore the old default at compile-time, start-time, or run-time.
• Reduce the default number of pages initially allocated for the pagecache from 100 to 20, for improved performance.
• Perform some UPDATE operations in a single pass instead of in two passes.
• Enhance the session extension to support WITHOUT ROWID tables.
• Fixed performance problems and potential stack overflows when creating views from multi-row VALUES clauses with hundreds of thousands of rows.
• Added the sha1 extension.
• Other performance improvements. Uses about 6.5% fewer CPU cycles.

Bug Fixes:

• Throw an error if the ON clause of a LEFT JOIN references tables to the right of the ON clause. This is the same behavior as PostgreSQL. Formerly, SQLite silently converted the LEFT JOIN into an INNER JOIN.
• Use the correct affinity for columns of automatic indexes.
• Ensure that the sqlite3_blob_reopen interface can correctly handle short rows.

• Fix the REPLACE statement for WITHOUT ROWID tables that lack secondary indexes so that it works correctly with triggers and foreign keys. This was a new bug caused by performance optimizations added in DISQLite3 5.17.0.
• Fix the sqlite3_value_text interface so that it correctly translates content generated by zeroblob() into a string of all 0x00 characters. This is a long-standing issue discovered after the DISQLite3 5.17.0 release.
• Fix the bytecode generator to deal with a subquery in the FROM clause that is itself a UNION ALL where one side of the UNION ALL is a view that contains an ORDER BY. This is a long-standing issue that was discovered after the release of DISQLite3 5.17.0.
• Adjust the sqlite3_column_count API so it more often returns the same values for PRAGMA statements as it did in prior releases, to minimize disruption to applications that might be using that interface in unexpected ways.

• Uses 9% fewer CPU cycles. (See the CPU performance measurement report for details on how this performance increase was computed.)
• Added experimental support for PRAGMA functions.
• Enhance the date and time functions so that the 'unixepoch' modifier works for the full span of supported dates.
• Changed the default configuration of the lookaside memory allocator from 500 slots of 128 bytes each into 125 slots of 512 bytes each.
• Enhanced the LIKE and GLOB matching algorithm to be faster for cases when the pattern contains multiple wildcards.
• Added the remember(V,PTR) SQL function as a loadable extension.

Bug Fixes:

• Fix a long-standing bug in the query planner that caused incorrect results on a LEFT JOIN where the left-hand table is a subquery and the join constraint is a bare column name coming from the left-hand subquery.
• Correctly handle the integer literal -0x8000000000000000 in the query planner.
• Fix a bug concerning the use of row values within triggers.

• Bug fixes to the row value logic that was introduced in version 3.15.0.
• Fix a NULL pointer dereference in ATTACH/DETACH following a maliciously constructed syntax error.
• Fix a crash that can occur following an out-of-memory condition in the built-in instr() function.
• In the JSON extension, fix the JSON validator so that it correctly rejects invalid backslash escapes within strings.

• Added the SQLITE_DBCONFIG_NO_CKPT_ON_CLOSE option to sqlite3_db_config.
• Enhanced “WHERE x NOT NULL” partial indexes so that they are usable if the “x” column appears in a LIKE or GLOB operator that has not been overloaded.
• Enhanced sqlite3_interrupt so that it interrupts checkpoint operations that are in process.
• Added SQLITE_FCNTL_WIN32_GET_HANDLE file control opcode.

Bug Fixes:

• Fix the VACUUM command so that it spills excess content to disk rather than holding everything in memory, and possible causing an out-of-memory error for larger database files.
• Fix a case where OR-connected terms in the ON clause of a LEFT JOIN might cause incorrect results.
• Fix a case where the use of row values in the ON clause of a LEFT JOIN might cause incorrect results.

• Added support for row values.
• Allow deterministic SQL functions in the WHERE clause of a partial index.
• Added support for SQLITE_DBCONFIG_MAINDBNAME.
• Added the ability to VACUUM an ATTACH-ed database.
• Enhance the fts5vocab virtual table to handle “ORDER BY term” efficiently.
• Miscellaneous micro-optimizations reduce CPU usage by more than 7% on common workloads. Most optimization in this release has been on the front-end (sqlite3_prepare_v2).

Bug Fixes:

• The multiply operator now correctly detects 64-bit integer overflow and promotes to floating point in all corner-cases.
• Correct handling of columns with redundant unique indexes when those columns are used on the LHS of an IN operator.
• Skip NULL entries on range queries in indexes on expressions.
• Ensure that the AUTOINCREMENT counters in the sqlite_sequence table are initialized doing “Xfer Optimization” on “INSERT … SELECT” statements.
• Make sure the ORDER BY LIMIT optimization works with IN operators on INTEGER PRIMARY KEYs.

• Added support for WITHOUT ROWID virtual tables.
• Improved the query planner so that the OR optimization can be used on virtual tables even if one or more of the disjuncts use the LIKE, GLOB, REGEXP, MATCH operators.
• Added the CSV virtual table for reading RFC 4180 formatted comma-separated value files.
• Added the carray() table-valued function extension.
• Enabled persistent loadable extensions using the new SQLITE_OK_LOAD_PERMANENTLY return code from the extension entry point.
• Added the SQLITE_DBSTATUS_CACHE_USED_SHARED option to sqlite3_db_status.
• Add the vfsstat loadable extension – a VFS shim that measures I/O together with an eponymous virtual table that provides access to the measurements.
• Improved algorithm for running queries with both an ORDER BY and a LIMIT where only the inner-most loop naturally generates rows in the correct order.
• The PRAGMA compile_options command now attempts to show the version number of the compiler that generated the library.
• Enhance PRAGMA table_info so that it provides information about eponymous virtual tables.
• Added the “win32-none” VFS, analogous to the “unix-none” VFS, that works like the default “win32” VFS except that it ignores all file locks.
• The query planner uses a full scan of a partial index instead of a full scan of the main table, in cases where that makes sense.
• Allow table-valued functions to appear on the right-hand side of an IN operator.
• Added two new interfaces: sqlite3_expanded_sql and sqlite3_trace_v2.
• Added the json_quote() SQL function to the json1 extension.
• Disable the authorizer callback while reparsing the schema.
• A performance enhancement to the page-cache “truncate” operation reduces COMMIT time by dozens of milliseconds on systems with a large page cache.

Bug Fixes:

• Fix the ALTER TABLE command so that it does not corrupt descending indexes when adding a column to a legacy file format database.
• Fix a NULL-pointer dereference/crash that could occurs when a transitive WHERE clause references a non-existent collating sequence.
• Improved the cost estimation for an index scan which includes a WHERE clause that can be partially or fully evaluated using columns in the index and without having to do a table lookup.
• Fix the sqlite3_trace_v2 interface so that it is disabled if either the callback or the mask arguments are zero, in accordance with the documentation.
• Correct affinity computations for a SELECT on the RHS of an IN operator.
• The ORDER BY LIMIT optimization is not valid unless the inner-most IN operator loop is actually used by the query plan.
• Fix an internal code generator problem that was causing some DELETE operations to no-op.

• Postpone I/O associated with TEMP files for as long as possible, with the hope that the I/O can ultimately be avoided completely.
• Added the sqlite3_db_config(db,SQLITE_DBCONFIG_ENABLE_LOAD_EXTENSION) interface which allows the sqlite3_load_extension C-API to be enabled while keeping the load_extension() SQL function disabled for security.

Bug fixes:

• Fix a problem with the multi-row one-pass DELETE optimization that was causing it to compute incorrect answers with a self-referential subquery in the WHERE clause.
• Fix a possible segfault with DELETE when table is a rowid table with an INTEGER PRIMARY KEY and the WHERE clause contains a OR and the table has one or more indexes that are able to trigger the OR optimization, but none of the indexes reference any table columns other than the INTEGER PRIMARY KEY.
• When checking for the WHERE-clause push-down optimization, verify that all terms of the compound inner SELECT are non-aggregate, not just the last term. Fix for ticket f7f8c97e97597.
• Fix a locking race condition in Windows that can occur when two or more processes attempt to recover the same hot journal at the same time.
• Fix a bug that can cause the DELETE operation to miss rows if PRAGMA reverse_unordered_selects is turned on.
• Fix a bug in the code generator that can cause incorrect results if two or more virtual tables are joined and the virtual table used in outer loop of the join has an IN operator constraint.
• Correctly interpret negative “PRAGMA cache_size” values when determining the cache size used for sorting large amounts of data.

• Support Delphi 10.1 Berlin Win32 and Win64.

• Add the Session Extension.
• Add the pre-update hook APIs.
• Enhanced the query planner so that the IS and IS NULL operators is able to drive indexes in a LEFT OUTER JOIN.
• Improved resistance against goofy query planner decisions caused by incomplete or incorrect modifications to the sqlite_stat1 table by the application.
• Fix a boundary condition error that can result in a crash during heavy SAVEPOINT usage.
• Fix views so that they inherit column datatypes from the table that they are defined against, when possible.
• Fix the query planner so that IS and IS NULL operators are able to drive an index on a LEFT OUTER JOIN.

• Added the sqlite3rbu_bp_progress interface to the RBU extension.
• Added the sqlite3_system_errno interface.

Potentially Disruptive Change:

• The default page size is increased from 1024 to 4096. The default cache size is changed from 2000 to -2000 so the same amount of cache memory is used by default. See the application note on the version 3.12.0 page size change for further information.

Backwards Compatibility:

• Because of continuing security concerns, the two-argument version of of the seldom-used and little-known fts3_tokenizer SQL function is disabled. The SQLITE_DBCONFIG_ENABLE_FTS3_TOKENIZER option to sqlite3_db_config allows to enable or disable it at run-time.

New features:

• The PRAGMA defer_foreign_keys=ON statement now also disables RESTRICT actions on foreign key.
• Added the FTS5 detail option.
• Added the “EXTRA” option to PRAGMA synchronous that does a sync of the containing directory when a rollback journal is unlinked in DELETE mode, for better durability.
• Add the SQLITE_FCNTL_JOURNAL_POINTER file control.
• Added support for LIKE, GLOB, and REGEXP operators on virtual tables.
• Added the colUsed field to sqlite3_index_info for use by the sqlite3_module.xBestIndex method.
• Enhance the PRAGMA cache_spill statement to accept a 32-bit integer parameter which is the threshold below which cache spilling is prohibited.
• Added the sqlite3_db_cacheflush interface.
• Added the sqlite3_strlike interface.
• Added the experimental sqlite3_snapshot_get, sqlite3_snapshot_open, and sqlite3_snapshot_free interfaces. These are subject to change or removal in a subsequent release.
• Added the json_group_array() and json_group_object SQL functions in the json extension.

Performance enhancements:

• The query planner considers the LIMIT clause when estimating the cost of ORDER BY.
• More efficient handling of application-defined SQL functions, especially in cases where the application defines hundreds or thousands of custom functions.
• Smaller and faster SQL parser.
• Only create statement journal files if two or more attached databases are (1) modified, (2) do not have PRAGMA synchronous set to OFF, and (3) do not have the journal_mode set to OFF, MEMORY, or WAL.
• Only create statement journal files when their size exceeds a threshold. Otherwise the journal is held in memory and no I/O occurs. The threshold can be configured at start-time using sqlite3_config(SQLITE_CONFIG_STMTJRNL_SPILL).
• The query planner is able to optimize IN operators on virtual tables even if the xBestIndex method does not set the sqlite3_index_constraint_usage.omit flag of the virtual table column to the left of the IN operator.
• The query planner now does a better job of optimizing virtual table accesses in a 3-way or higher join where constraints on the virtual table are split across two or more other tables of the join.
• Enhanced WAL mode so that it works efficiently with transactions that are larger than the cache_size.
• Enhanced the query planner so that it is able to use a covering index as part of the OR optimization.
• Avoid recomputing NOT NULL and CHECK constraints on unchanged columns in UPDATE statement.
• Reduce the amount of memory needed to hold the schema.
• Many small and micro-optimizations, resulting in a library that is faster than the previous release.

General improvements:

• Fix an FTS5 issue in which the 'optimize' command could cause index corruption.
• Fix a buffer overread that might occur if FTS5 is used to query a corrupt database file.
• Increase the maximum “scope” value for the spellfix1 extension from 6 to 30.
• When using memory-mapped I/O map the database file read-only so that stray pointers and/or array overruns in the application cannot accidently modify the database file.
• Enhance the 'utc' modifier in the date and time functions so that it is a no-op if the date/time is known to already be in UTC. (This is not a compatibility break since the behavior has long been documented as “undefined” in that case.)

Bug fixes:

• Make sure the sqlite3_set_auxdata values from multiple triggers within a single statement do not interfere with one another.
• Fix the code generator for expression of the form “x IN (SELECT…)” where the SELECT statement on the RHS is a correlated subquery.
• Fix an issue with incorrect sharing of VDBE temporary registers between co-routines that could cause incorrect query results in obscure cases.
• Fix a problem in the sqlite3_result_subtype interface that could cause problems for the json1 extension under obscure circumstances.
• Escape control characters in JSON strings.
• Fix a bug in the query planner that could generate incorrect results when a scalar subquery attempts to use the block sorting optimization.
• Fix inconsistent integer to floating-point comparison operations that could result in a corrupt index if the index is created on a table column that contains both large integers and floating point values of similar magnitude.
• Fix an infinite-loop in the query planner that could occur on malformed common table expressions.

• Fix the schema parser so that it interprets certain (obscure and ill-formed) CREATE TABLE statements the same as legacy.
• Fix a query planner problem that could result in an incorrect answer due to the use of automatic indexing in subqueries in the FROM clause of a correlated scalar subqueries.

• Added the sqlite3_value_subtype and sqlite3_result_subtype interfaces.
• Enhanced the dbstat virtual table so that it can be used as a table-valued function where the argument is the schema to be analyzed.
• Do not allow a comma at the end of a JSON array or object.
• Performance optimizations on the JSON parser.

This is a bug fix release:

• Do not allow rowid in a UNIQUE or PRIMARY KEY constraint.
• Do not allow table-valued function syntax to be used on a view.
• Fix a problem with FTS5 “ORDER BY rowid DESC” queries and large terms.
• Fix a crash in FTS5 caused by specifying NULL as the rhs of a MATCH operator.
• Fix a memory leak and off-by-one error in the json1 extension.
• Fix json_set() so that it can overwrite a value that was previously overwritten during the same call.
• Fix an obscure bug in the initial output of json_tree() when using a path to an object contained within an array.
• Remove old refactoring code from DISQLite3_Full_Text_Search_Form_Main.pas which caused a (harmless) runtime error.

New Features And Enhancements:

• Support Delphi 10 Seattle Win32 and Win64.
• The CREATE VIEW statement now accepts an optional list of column names following the view name.
• Added support for indexes on expressions.
• Added support for table-valued functions in the FROM clause of a SELECT statement.
• Added support for eponymous virtual tables.
• Added the json1 extension.
• A VIEW may now reference undefined tables and functions when initially created. Missing tables and functions are reported when the VIEW is used in a query.
• The query planner is now able to use partial indexes that contain AND-connected terms in the WHERE clause.

Other changes:

• Modify the FTS5 leaf page format to permit faster seek operations. This is a file-format change. Any existing databases can be upgraded by running the fts5 'rebuild' command.
• Modify the FTS5 custom tokenizer interface to permit synonym support. The fts5_api.iVersion value is now set to 2. Existing FTS5 custom tokenizers will need to be updated to use the new api version.
• The sqlite3_memory_alarm interface, which has been deprecated and undocumented for 8 years, is changed into a no-op.

Fixes:

• Added a memory barrier in the implementation of sqlite3_initialize to help ensure that it is thread-safe.
• Fix the OR optimization so that it always ignores subplans that do not use an index.
• Do not apply the WHERE-clause pushdown optimization on terms that originate in the ON or USING clause of a LEFT JOIN.

• Added the experimental FTS5 extension. Note that this extension is experimental and subject to change in incompatible ways.
• Added the sqlite3_value_dup and sqlite3_value_free interfaces.
• Added the sqlite3_result_zeroblob64 and sqlite3_bind_zeroblob64 interfaces.
• Enhance the spellfix1 extension to support ON CONFLICT clauses.
• The IS operator is now able to drive indexes.
• Enhance the query planner to permit automatic indexing on FROM-clause subqueries that are implemented by co-routine.
• Disallow the use of “rowid” in common table expressions.
• Added the PRAGMA cell_size_check command for better and earlier detection of database file corruption.
• Added the matchinfo 'b' flag to the matchinfo() function in FTS3.
• The sqlite3_profile callback is invoked (by sqlite3_reset or sqlite3_finalize) for statements that did not run to completion.
• Enhance the page cache so that it can preallocate a block of memory to use for the initial set page cache lines. Set the default preallocation to 100 pages. Yields about a 5% performance increase on common workloads.
• Miscellaneous micro-optimizations result in 22.3% more work for the same number of CPU cycles relative to the previous release. DISQLite now runs twice as fast as two years ago.
• Add TDISQLite3Database.BusyTimeout property.
• Improve sqlite3_exec_with_callback error results.

• Fix an index corruption issue introduced by DISQLite3 5.7.0. An index with a TEXT key could be corrupted by an INSERT into the corresponding table if the table had two nested triggers that converted the key value to INTEGER and back to TEXT again.

• Added the y format string to the matchinfo() function of FTS3.
• Add the dbstat virtual table.
• Performance improvements for ORDER BY, VACUUM, CREATE INDEX, PRAGMA integrity_check, and PRAGMA quick_check.
• Fix many obscure problems discovered while SQL fuzzing.

• Added the sqlite3_status64 interface.
• Fix memory size tracking so that it works even if SQLite uses more than 2GiB of memory.
• Added the PRAGMA index_xinfo command.
• Fix a potential 32-bit integer overflow problem in the sqlite3_blob_read and sqlite3_blob_write interfaces.
• Improve the performance of fts3/4 queries that use the OR operator and at least one auxiliary fts function.
• Fix a bug in the fts3 snippet() function causing it to omit leading separator characters from snippets that begin with the first token in a column.

• Add support for Delphi XE8 Win32 and Win64.

• Fix a bug that could lead to incorrect results if the qualifying constraint of a partial index appears in the ON clause of a LEFT JOIN.
• Improve the performance of fts3/4 queries that use the OR operator and at least one auxiliary fts function.

New Features:

• Added the PRAGMA data_version command that can be used to determine if a database file has been modified by another process.
• Added the SQLITE_CHECKPOINT_TRUNCATE option to the sqlite3_wal_checkpoint_v2 interface, with corresponding enhancements to PRAGMA wal_checkpoint. sqlite3_wal_checkpoint_v2(SQLITE_CHECKPOINT_TRUNCATE) truncates the WAL file even if there is no checkpoint work to be done.
• Added the sqlite3_stmt_scanstatus interface.
• The sqlite3_table_column_metadata is enhanced to work correctly on WITHOUT ROWID tables and to check for the existence of a table if the column name parameter is nil.
• Added the SQLITE_CONFIG_PMASZ start-time option.
• Added the SQLITE_CONFIG_PCACHE_HDRSZ option to sqlite3_config which makes it easier for applications to determine the appropriate amount of memory for use with SQLITE_CONFIG_PAGECACHE.
• The number of rows in a VALUES clause is no longer limited.
• Added the eval loadable extension that implements an eval() SQL function that will recursively evaluate SQL.

Performance Enhancements:

• Reduce the number of memcopy operations involved in balancing a b-tree, for 3.2% overall performance boost.
• Improvements to cost estimates for the skip-scan optimization.
• The automatic indexing optimization is now capable of generating a partial index if that is appropriate.

Bug fixes:

• Ensure durability following a power loss with “PRAGMA journal_mode=TRUNCATE” by calling fsync() right after truncating the journal file.
• The query planner now recognizes that any column in the right-hand table of a LEFT JOIN can be NULL, even if that column has a NOT NULL constraint. Avoid trying to optimize out NULL tests in those cases.
• Make sure ORDER BY puts rows in ascending order even if the DISTINCT operator is implemented using a descending index.
• Fix data races that might occur under stress when running with many threads in shared cache mode where some of the threads are opening and closing connections.
• Fix a bug in the sorting logic that can cause output to appear in the wrong order on queries that contains an ORDER BY clause, a LIMIT clause, and that have approximately 60 or more columns in the result set.
• Fix obscure crash bugs.

• Bug fix: Ensure the cached KeyInfo objects (an internal abstraction not visible to the application) do not go stale when operating in shared cache mode and frequently closing and reopening some database connections while leaving other database connections on the same shared cache open continuously.
• Bug fix: Recognize that any column in the right-hand table of a LEFT JOIN can be NULL even if the column has a NOT NULL constraint. Do not apply optimizations that assume the column is never NULL.

• Enhance the ROLLBACK command so that pending queries are allowed to continue as long as the schema is unchanged. Formerly, a ROLLBACK would cause all pending queries to fail with an SQLITE_ABORT or SQLITE_ABORT_ROLLBACK error. That error is still returned if the ROLLBACK modifies the schema.
• Bug fix: Make sure that NULL results from OP_Column are fully and completely NULL and do not have the MEM_Ephem bit set.
• Bug fix: The %c format in sqlite3_mprintf is able to handle precisions greater than 70.
• Bug fix: Do not automatically remove the DISTINCT keyword from a SELECT that forms the right-hand side of an IN operator since it is necessary if the SELECT also contains a LIMIT.

Performance Enhancements:

• Many micro-optimizations result in 20.3% more work for the same number of CPU cycles relative to the previous release. The cumulative performance increase since DISQLite3 5.0.0 is 61%. Your performance may vary.
• The sorter can use auxiliary helper threads to increase real-time response. This feature is off by default and may be enabled using the PRAGMA threads command. DISQLite3 Professional only.
• Enhance the skip-scan optimization so that it is able to skip index terms that occur in the middle of the index, not just as the left-hand side of the index.
• Improved optimization of CAST operators.
• Various improvements in how the query planner uses sqlite_stat4 information to estimate plan costs.

New Features:

• Added new interfaces with 64-bit length parameters:
  • sqlite3_malloc64
  • sqlite3_realloc64
  • sqlite3_bind_blob64
  • sqlite3_result_blob64
  • sqlite3_bind_text64 and
  • sqlite3_result_text64.
• Added the new interface sqlite3_msize that returns the size of a memory allocation obtained from sqlite3_malloc64 and its variants.
• Added the SQLITE_LIMIT_WORKER_THREADS option to sqlite3_limit and PRAGMA threads command for configuring the number of available worker threads.
• The spellfix1 extension allows the application to optionally specify the rowid for each INSERT.

Bug Fixes:

• Fix a bug in the partial index implementation that might result in an incorrect answer if a partial index is used in a subquery or in a view.
• Fix a query planner bug that might cause a table to be scanned in the wrong direction (thus reversing the order of output) when a DESC index is used to implement the ORDER BY clause on a query that has an identical GROUP BY clause.
• Fix a bug in sqlite3_trace that was causing it to sometimes fail to print an SQL statement if that statement needed to be re-prepared.
• In PRAGMA journal_mode=TRUNCATE mode, call fsync() immediately after truncating the journal file to ensure that the transaction is durable across a power loss.
• Fix a couple of problems associated with running an UPDATE or DELETE on a VIEW with a rowid in the WHERE clause.

• Support Delphi XE7 Win32 and Win64.

• Added support for hexadecimal integer literals in the SQL parser. (Ex: 0x123abc)
• Enhanced the PRAGMA integrity_check command to detect UNIQUE and NOT NULL constraint violations.
• Increase the maximum number of attached databases from 62 to 125.
• Increase the timeout in WAL mode before issuing an SQLITE_PROTOCOL error from 1 second to 10 seconds.
• Added the likely(X) SQL function.
• Trigger automatic reprepares on all prepared statements when ANALYZE is run.

Performance Enhancements:

• Deactivate the DISTINCT keyword on subqueries on the right-hand side of the IN operator.
• Add the capability of evaluating an IN operator as a sequence of comparisons as an alternative to using a table lookup. Use the sequence of comparisons implementation in circumstances where it is likely to be faster, such as when the right-hand side of the IN operator is small and/or changes frequently.
• The query planner now uses sqlite_stat4 information (created by ANALYZE) to help determine if the skip-scan optimization is appropriate.
• Ensure that the query planner never tries to use a self-made transient index in place of a schema-defined index.
• Other minor tweaks to improve the quality of VDBE code.

Bug Fixes:

• Fix a bug in CREATE UNIQUE INDEX, introduced when WITHOUT ROWID support was added, that allows a non-unique NOT NULL column to be given a UNIQUE index.
• Fix a bug in R-Tree extension, introduced in the previous release, that can cause an incorrect results for queries that use the rowid of the R-Tree on the left-hand side of an IN operator.
• Fix the sqlite3_stmt_busy interface so that it gives the correct answer for ROLLBACK statements that have been stepped but never reset.
• Fix a bug in that would cause a null pointer to be dereferenced if a column with a DEFAULT that is an aggregate function tried to usee its DEFAULT.
• Fix a column affinity problem with the IN operator.
• Fix the ANALYZE command so that it adds correct samples for WITHOUT ROWID tables in the sqlite_stat4 table.

• Added support for partial sorting by index.
• Enhance the query planner so that it always prefers an index that uses a superset of WHERE clause terms relative to some other index.
• Improvements to the automerge command of FTS4 to better control the index size for a full-text index that is subject to a large number of updates.
• Added the sqlite3_rtree_query_callback interface to R-Tree extension
• Added new URI query parameters “nolock” and “immutable”.
• Use less memory by not remembering CHECK constraints on read-only database connections.
• Enable the OR optimization for WITHOUT ROWID tables.
• Render expressions of the form “x IN (?)” (with a single value in the list on the right-hand side of the IN operator) as if they where “x==?”, Similarly optimize “x NOT IN (?)”
• Added the SQLITE_IOCAP_IMMUTABLE bit to the set of bits that can be returned by the xDeviceCharacteristics method of a VFS.

Bug Fixes:

• OFFSET clause ignored on queries without a FROM clause.
• Assertion fault on queries involving expressions of the form “x IN (?)”.
• Incorrect column datatype reported.
• Duplicate row returned on a query against a table with more than 16 indices, each on a separate column, and all used via OR-connected constraints.
• Partial index causes assertion fault on UPDATE OR REPLACE.
• Crash when calling undocumented SQL function sqlite_rename_parent() with NULL parameters.
• ORDER BY ignored if the query has an identical GROUP BY.
• The group_concat(x,'') SQL function returns NULL instead of an empty string when all inputs are empty strings.
• Fix a bug in the VDBE code generator that caused crashes when doing an INSERT INTO … SELECT statement where the number of columns being inserted is larger than the number of columns in the destination table.
• Fix a problem in FTS4 where the left-most column that contained the notindexed column name as a prefix was not indexed rather than the column whose name matched exactly.
• Fix the sqlite3_db_readonly interface so that it returns true if the database is read-only due to the file format write version number being too large.

• Support Delphi XE6 Win32 and Win64.
• New functions QuoteIdent and QuoteIdent16. Very much like QuotedStr and QuotedStr16 but they also quote SQLite keywords.
• SqlBufToFloat and related functions:
  • Length parameter is now an Integer type.
  • Process trailing whitespace and control characters similar to SysUtils.StrToFloat.
  • SqlBufToFloatDef and SqlStrToFloatDef functions must successfully process the entire string, otherwise the default value is returned.
• The DISQLite3_Full_Text_Search demo adjusts line breaks before displaying the file's text. This solves display problems with Mac and Unix files.

• Add a fix for a problem that might cause incorrect query results on a query that mixes DISTINCT, GROUP BY in a subquery, and ORDER BY.

• Fix a potential buffer overread that could result when trying to search a corrupt database file.

• Add TDISQLite3Database design-time properties: TDISQLite3Database.Access, TDISQLite3Database.Cache, TDISQLite3Database.Mutex, TDISQLite3Database.Options, TDISQLite3Database.VFS.
• New TDISQLite3Database.ReadOnly and TDISQLite3Database.ReadOnly16 methods.
• Apply the SQLITE_DETERMINISTIC optimization when creating these functions: math, regexp, regextract and zlib.
• Code optimization and refactoring for improved performance.
• When computing the cost of the skip-scan optimization, take into account the fact that multiple seeks are required.
• Bug fix: Fix the char() SQL function so that it returns an empty string rather than an “out of memory” error when called with zero arguments.
• Bug fix: DISTINCT now recognizes that a zeroblob and a blob of all 0x00 bytes are the same thing.
• Bug fix: Compute the correct answer for queries that contain an IS NOT NULL term in the WHERE clause and also contain an OR term in the WHERE clause.
• Bug fix: Make sure “rowid” columns are correctly resolved in joins between normal tables and WITHOUT ROWID tables.
• Bug fix: Make sure the same temporary registers are not used in concurrent co-routines used to implement compound SELECT statements containing ORDER BY clauses, as such use can lead to incorrect answers.
• Bug fix: Ensure that “ORDER BY random()” clauses do not get optimized out.
• Bug fix: Repair a name-resolution error that can occur in sub-select statements contained within a TRIGGER.
• Bug fix: Fix column default values expressions of the form “DEFAULT(-(-9223372036854775808))” so that they work correctly, initializing the column to a floating point value approximately equal to +9223372036854775808.0.
• Bug fix: sqlite3_is_memory_name is now case sensitive, the same way SQLite handles the :memory: file name.

• Fix a bug hat causes rows to go missing on some queries with OR clauses and IS NOT NULL operators in the WHERE clause.

• Add support for common table expressions. It is now possible to write a single SELECT statement that will query a tree or graph, using either a depth-first or a breadth-first search. As part of this change, SQLite now accepts a VALUES clause anyplace that a SELECT statement is valid.
• Enhancements:
  • Added the printf() SQL function.
  • Added SQLITE_DETERMINISTIC as an optional bit in the 4th argument to the sqlite3_create_function and related interfaces, providing applications with the ability to create new functions that can be factored out of inner loops when they have constant arguments.
  • Add SQLITE_READONLY_DBMOVED error code, returned at the beginning of a transaction, to indicate that the underlying database file has been renamed or moved out from under SQLite.
  • Allow arbitrary expressions, including function calls and subqueries, in the filename argument to ATTACH.
  • Reseed the PRNG used by sqlite3_randomness(N,P) when invoked with N=0.
  • Enhance the spellfix1 virtual table so that it can search efficiently by rowid.
• Many small performance enhancements which should give a small speed boost to legacy applications.

• Support for WITHOUT ROWID tables. This is a significant extension to SQLite. Database files that contain WITHOUT ROWID tables are not readable or writable by prior versions of SQLite, however databases that do not use WITHOUT ROWID tables are fully backwards and forwards compatible.
• Potential incompatibility change: In all prior versions of SQLite, a cast from a very large positive floating point number into an integer resulted in the most negative integer. In other words, CAST(+99.9e99 to INT) would yield -9223372036854775808. This behavior came about because it is what x86/x64 hardware does for the equivalent cast in the C language. But the behavior is bizarre. And so it has been changed effective with this release so that a cast from a floating point number into an integer returns the integer between the floating point value and zero that is closest to the floating point value. Hence, CAST(+99.9e99 to INT) now returns +9223372036854775807. Since routines like sqlite3_column_int64 do an implicit cast if the value being accessed is really a floating point number, they are also affected by this change.
• Performance enhancements: The skip-scan optimization is now available for databases that have been processed by ANALYZE. Constant SQL functions are now factored out of inner loops, which can result in a significant speedup for queries that contain WHERE clause terms like “date>datetime('now','-2 days')”. And various high-runner internal routines have been refactored for reduced CPU load.
• Extended the virtual table interface, and in particular the sqlite3_index_info object to allow a virtual table to report its estimate on the number of rows that will be returned by a query.
• Update the R-Tree extension to make use of the enhanced virtual table interface.

• Added the unlikely() and likelihood() SQL functions to be used as hints to the query planner.
• Enhancements to the query planner:
  1. Take into account the fact WHERE clause terms that cannot be used with indices still probably reduce the number of output rows.
  2. Estimate the sizes of table and index rows and use the smallest applicable B-Tree for full scans and “count(*)” operations.
• Added the soft_heap_limit pragma.
• Added support for the sqlite_stat4 table. This is an enhancement over the sqlite_stat3 table which only recorded histogram data for the left-most column of each index whereas sqlite_stat4 tables record histograph data from all columns of each index. sqlite_stat3 tables might still read but are no longer written.
• Added support for “sz=NNN” parameters at the end of sqlite_stat1.stat fields used to specify the average length in bytes for table and index rows.
• Avoid running foreign-key constraint checks on an UPDATE if none of the modified columns are associated with foreign keys.
• Added the win32-longpath VFS, permitting filenames up to 32K characters in length (DISQLite3 Pro only).
• The Date And Time Functions are enhanced so that the current time (ex: julianday('now')) is always the same for multiple function invocations within the same sqlite3_step call.
• Add the “totype” extension, implementing the tointeger() and toreal() SQL functions.
• FTS4 queries are better able to make use of docid < limit constraints to limit the amount of I/O required.
• The VACUUM command packs the database about 1% tighter.
• Refactor the implementation of PRAGMA statements to improve parsing performance.
• Bug fix: Return the correct answer for “SELECT count(*) FROM table” even if there is a partial index on the table.

• Support Delphi XE5 Win32 and Win64.

• Fix a bug in the optimization that attempts to omit unused LEFT JOINs.

• Fix a query planner bug involving a LEFT JOIN with a BETWEEN or LIKE/GLOB constraint and then another INNER JOIN to the right that involves an OR constraint.
• Fix a query planner bug that could result in a segfault when querying tables with a UNIQUE or PRIMARY KEY constraint with more than four columns.

• Add support for partial indexes.
• Cut-over to the next generation query planner for faster and better query plans.
• The EXPLAIN QUERY PLAN output no longer shows an estimate of the number of rows generated by each loop in a join.
• Added the FTS4 notindexed option, allowing non-indexed columns in an FTS4 table.
• Added the SQLITE_STMTSTATUS_VM_STEP option to sqlite3_stmt_status.
• Added the cache_spill pragma.
• Added the query_only pragma.
• Added the defer_foreign_keys pragma and the sqlite3_db_status(db, SQLITE_DBSTATUS_DEFERRED_FKS,…) interface.
• Added the “percentile()” function as a loadable extension.
• Add the sqlite3_cancel_auto_extension(X) interface.
• A running SELECT statement that lacks a FROM clause (or any other statement that never reads or writes from any database file) will not prevent a read transaction from closing.
• Issue an SQLITE_WARNING_AUTOINDEX warning on the SQLITE_CONFIG_LOG whenever the query planner uses an automatic index.
• Added an optional 5th parameter defining the collating sequence to the “next_char()” extension SQL function.
• The SQLITE_BUSY_SNAPSHOT extended error code is returned in WAL mode when a read transaction cannot be upgraded to a write transaction because the read is on an older snapshot.
• Allow read transactions to be freely opened and closed by SQL statements run from within the implementation of application-defined SQL functions if the function is called by a SELECT statement that does not access any database table.
• Bug fix: In the unicode61 tokenizer of FTS4, treat all private code points as identifier symbols.
• Bug fix: Bare identifiers in ORDER BY clauses bind more tightly to output column names, but identifiers in expressions bind more tightly to input column names. Identifiers in GROUP BY clauses always prefer output column names, however.
• Bug fixes: Multiple problems in the legacy query optimizer were fixed by the move to NGQP.

• Support Delphi XE4 Win32 and Win64.
• Add support for memory-mapped I/O.
• Add the sqlite3_strglob convenience interface.
• Assigned the integer at offset 68 in the database header as the Application ID for when SQLite is used as an application file-format. Added the PRAGMA application_id command to query and set the Application ID.
• Report rollback recovery in the error log as SQLITE_NOTICE_RECOVER_ROLLBACK. Change the error log code for WAL recover from SQLITE_OK to SQLITE_NOTICE_RECOVER_WAL.
• Report the risky uses of unlinked database files and database filename aliasing as SQLITE_WARNING messages in the error log.
• Enhance the extension loading mechanism to be more flexible (while still maintaining backwards compatibility) in two ways:
  1. If the default entry point “sqlite3_extension_init” is not present in the loadable extension, also try an entry point “sqlite3_X_init” where “X” is based on the shared library filename. This allows every extension to have a different entry point, which allows them to be statically linked with no code changes.
  2. The shared library filename passed to sqlite3_load_extension may omit the filename suffix, and an appropriate architecture-dependent suffix (“.dll”) will be added automatically.
• Added many new loadable extensions, including amatch, closure, fuzzer, ieee754, and nextchar.
• Enhance FTS3 to avoid using excess stack space when there are a huge number of terms on the right-hand side of the MATCH operator. A side-effect of this change is that the MATCH operator can only accommodate 12 NEAR operators at a time.
• Enhance the fts4aux virtual table so that it can be a TEMP table.
• Added the fts3tokenize virtual table to the full-text search logic.
• Query planner enhancement: Use the transitive property of constraints to move constraints into the outer loops of a join whenever possible, thereby reducing the amount of work that needs to occur in inner loops.
• Bug fix: Fix a potential database corruption bug in shared cache mode when one database connection is closed while another is in the middle of a write transaction.
• Bug fix: Only consider AS names from the result set as candidates for resolving identifiers in the WHERE clause if there are no other matches. In the ORDER BY clause, AS names take priority over any column names.
• Bug fix: Do not allow a virtual table to cancel the ORDER BY clause unless all outer loops are guaranteed to return no more than one row result.
• Bug fix: Do not suppress the ORDER BY clause on a virtual table query if an IN constraint is used.
• Bug fix: Make sure PRAGMA statements appear in sqlite3_trace output.
• Bug fix: When a compound query that uses an ORDER BY clause with a COLLATE operator, make sure that the sorting occurs according to the specified collation and that the comparisons associate with the compound query use the native collation.
• Bug fix: Makes sure the authorizer callback gets a valid pointer to the string “ROWID” for the column-name parameter when doing an UPDATE that changes the rowid.
• Bug fix: Do not move WHERE clause terms inside OR expressions that are contained within an ON clause of a LEFT JOIN.
• Bug fix: Make sure an error is always reported when attempting to preform an operation that requires a collating sequence that is missing.

• Fix a bug that could result in database corruption two or more processes try to access the same database file at the same time and immediately after third process crashed in the middle of committing to that same file.

• Fix for a bug in the ORDER BY optimizer that was introduced in DISQLite3 4.1.2 which would sometimes optimize out the sorting step when in fact the sort was required.
• Fix a long-standing bug in the CAST expression that would recognize UTF16 characters as digits even if their most-significant-byte was not zero.
• Fix a bug in the NEAR operator of FTS3 when applied to subfields.
• Fix a long-standing bug in the storage engine that would (very rarely) cause a spurious report of an SQLITE_CORRUPT error but which was otherwise harmless.
• Added the PRAGMA foreign_key_check command.
• Added new extended error codes for all SQLITE_CONSTRAINT errors
• Added the SQLITE_READONLY_ROLLBACK extended error code for when a database cannot be opened because it needs rollback recovery but is read-only.
• Added SQL functions unicode(A) and char(X1,…,XN)].
• Performance improvements for PRAGMA incremental_vacuum, especially in cases where the number of free pages is greater than what will fit on a single trunk page of the freelist.
• Improved optimization of queries containing aggregate min() or max().
• Enhance virtual tables so that they can potentially use an index when the WHERE clause contains the IN operator.
• Allow indices to be used for sorting even if prior terms of the index are constrained by IN operators in the WHERE clause.
• Enhance the PRAGMA table_info command so that the “pk” column is an increasing integer to show the order of columns in the primary key.
• Enhance the query optimizer to exploit transitive join constraints.
• Performance improvements in the query optimizer.
• Allow the error message from PRAGMA integrity_check to be longer than 20000 bytes.
• Improved name resolution for deeply nested queries.
• Improved error messages in the RTREE extension.
• Improved error messages for “foreign key mismatch” showing the names of the two tables involved.
• Enhance the spellfix1 extension so that the edit distance cost table can be changed at runtime by inserting a string like 'edit_cost_table=TABLE' into the “command” field.
• Bug fix: repair a long-standing problem that could cause incorrect query results in a 3-way or larger join that compared INTEGER fields against TEXT fields in two or more places.
• Bug fix: Issue an error message if the 16-bit reference counter on a view overflows due to an overly complex query.
• Bug fix: Avoid leaking memory on LIMIT and OFFSET clauses in deeply nested UNION ALL queries.
• Bug fix: Make sure the schema is up-to-date prior to running pragmas table_info, index_list, index_info, and foreign_key_list.
• DISQLite3Database: Remove deprecated functions and procedures.
• DISQLite3Database: User helper functions and procedures from DIUtils.pas to remove duplicates.

• Fix a bug that causes an ORDER BY clause to be optimized out of a three-way join when the ORDER BY is actually required.

• Fix a bug, introduced in DISQLite3 4.1.0, that causes a segfault if the AS name of a result column of a SELECT statement is used as a logical term in the WHERE clause.

• Added the sqlite3_errstr interface.
• Avoid invoking the sqlite3_trace callback multiple times when a statement is automatically reprepared due to SQLITE_SCHEMA errors.
• Enhance IN operator processing to make use of indices with numeric affinities.
• Do full-table scans using covering indices when possible, under the theory that an index will be smaller and hence can be scanned with less I/O.
• Enhance the query optimizer so that ORDER BY clauses are more aggressively optimized, especially in joins where various terms of the ORDER BY clause come from separate tables of the join.
• Add the ability to implement FROM clause subqueries as coroutines rather that manifesting the subquery into a temporary table.
• Added the busy_timeout pragma.
• Added the instr SQL function.
• Added the SQLITE_FCNTL_BUSYHANDLER file control, used to allow VFS implementations to get access to the busy handler callback.
• The xDelete method in the built-in VFSes now return SQLITE_IOERR_DELETE_NOENT if the file to be deleted does not exist.
• Bug fix: Avoid various concurrency problems in shared-cache mode.
• Bug fix: Avoid a deadlock or crash if the backup API, shared cache, and the Encryption Extension are all used at once.
• Bug fix: Fix a 32-bit overflow problem on CREATE INDEX for databases larger than 16GB.
• Bug fix: Avoid segfault when using the COLLATE operator inside of a CHECK constraint or view in shared cache mode.

• Support Delphi XE3 Win32 and Win64.
• Add the sqlite3_close_v2 interface.
• Enhancements to PRAGMA integrity_check and PRAGMA quick_check so that they can optionally check just a single attached database instead of all attached databases.
• Enhancements to WAL mode processing that ensure that at least one valid read-mark is available at all times, so that read-only processes can always read the database.
• Performance enhancements in the sorter used by ORDER BY and CREATE INDEX.
• Better handling of aggregate queries where the aggregate functions are contained within subqueries.
• Enhance the query planner so that it tries to use a covering index on queries that make use of or optimization.

• In-memory databases that are specified using URI filenames are allowed to use shared cache, so that the same in-memory database can be accessed from multiple database connections.
• Recognize and use the mode=memory query parameter in URI filenames.
• Avoid resetting the schema of shared cache connections when any one connection closes. Instead, wait for the last connection to close before reseting the schema.
• In the RTREE extension, when rounding 64-bit floating point numbers to 32-bit for storage, always round in a direction that causes the bounding box to get larger.
• The sqlite3_create_module and sqlite3_create_module_v2 interfaces return SQLITE_MISUSE on any attempt to overload or replace a virtual table module. The destructor is always called in this case, in accordance with historical and current documentation.
• New FTS3 / FTS4 unicode61 tokenizer with full unicode case folding according to rules in Unicode Version 6.1 and recognizes unicode space and punctuation characters and uses those to separate tokens.

• Fix a bug introduced in DISQLite3 3.1.4 that can cause a segfault for certain obscure nested aggregate queries.

• Add the SQLITE_DBSTATUS_CACHE_WRITE option for sqlite3_db_status.
• Optimize the typeof() and length() SQL functions so that they avoid unnecessary reading of database content from disk.
• Add the FTS4 “merge” command, the FTS4 “automerge” command, and the FTS4 “integrity-check” command.
• Report the name of specific CHECK constraints that fail.
• Make use of OVERLAPPED in the windows VFS to avoid some system calls and thereby obtain a performance improvement.
• More aggressive optimization of the AND operator when one side or the other is always false.
• Improved performance of queries with many OR-connected terms in the WHERE clause that can all be indexed.
• Add the SQLITE_RTREE_INT_ONLY compile-time option to force the R*Tree Extension Module to use integer instead of floating point values for both storage and computation (DISQLite3 Source Code edition only).
• Enhance the PRAGMA integrity_check command to use much less memory when processing multi-gigabyte databases.
• Allow virtual table constructors to be invoked recursively.
• Improved optimization of ORDER BY clauses on compound queries.
• Improved optimization of aggregate subqueries contained within an aggregate query.
• Bug fix: Fix the RELEASE command so that it does not cancel pending queries. This repairs a problem introduced in DISQLite3 3.1.3.
• Bug fix: Do not discard the DISTINCT as superfluous unless a subset of the result set is subject to a UNIQUE constraint and it none of the columns in that subset can be NULL.
• Do not optimize away an ORDER BY clause that has the same terms as a UNIQUE index unless those terms are also NOT NULL.
• Add TParam filtering to the DISQLite3_ClientDataSet_Grid demo project.

• Enhance the INSERT syntax to allow multiple rows to be inserted via the VALUES clause.
• Enhance the CREATE VIRTUAL TABLE command to support the IF NOT EXISTS clause.
• Added the sqlite3_stricmp interface as a counterpart to sqlite3_strnicmp.
• Added the sqlite3_db_readonly interface.
• Added the SQLITE_FCNTL_PRAGMA file control, giving VFS implementations the ability to add new PRAGMA statements or to override built-in PRAGMAs.
• Queries of the form: “SELECT max(x), y FROM table” returns the value of y on the same row that contains the maximum x value.
• Added support for the FTS4 languageid option.
• Documented support for the FTS4 content option. This feature has actually been in the code since DISQLite3 3.1.1 but is only now considered to be officially supported.
• Pending statements no longer block ROLLBACK. Instead, the pending statement will return SQLITE_ABORT upon next access after the ROLLBACK.
• Fix a bug introduced in SQLite3 3.7.10 that might cause a LEFT JOIN to be incorrectly converted into an INNER JOIN if the WHERE clause indexable terms connected by OR.

• The default schema format number is changed from 1 to 4. This means that, unless the PRAGMA legacy_file_format=ON statement is run, newly created database files will be unreadable by version of SQLite prior to 3.3.0 (2006-01-10), DISQLite3 1.2.4 (2005-12-28). It also means that the descending indices are enabled by default.
• The sqlite3_pcache_methods structure and the SQLITE_CONFIG_PCACHE and SQLITE_CONFIG_GETPCACHE configuration parameters are deprecated. They are replaced by a new sqlite3_pcache_methods2 structure and SQLITE_CONFIG_PCACHE2 and SQLITE_CONFIG_GETPCACHE2 configuration parameters.
• Added the powersafe overwrite property to the VFS interface. Provide the SQLITE_IOCAP_POWERSAFE_OVERWRITE I/O capability, the SQLITE_POWERSAFE_OVERWRITE compile-time option, and the “psow=BOOLEAN” query parameter for URI filenames.
• Added the sqlite3_db_release_memory interface and the shrink_memory pragma. Added the sqlite3_db_filename interface.
• Added the sqlite3_stmt_busy interface.
• Added the sqlite3_uri_boolean and sqlite3_uri_int64 interfaces.
• If the argument to PRAGMA cache_size is negative N, that means to use approximately -1024*N bytes of memory for the page cache regardless of the page size.
• Enhanced the query planner to support index queries with range constraints on the rowid.
• Enhanced the query planner flattening logic to allow UNION ALL compounds to be promoted upwards to replace a simple wrapper SELECT even if the compounds are joins.
• Enhanced the query planner so that the xfer optimization can be used with INTEGER PRIMARY KEY ON CONFLICT as long as the destination table is initially empty.
• Enhanced the windows VFS so that all system calls can be overridden using the xSetSystemCall interface.
• Change the windows VFS to report a sector size of 4096 instead of the old default of 512.
• Bug fix: Add an additional xSync when restarting a WAL in order to prevent an exceedingly unlikely but theoretically possible database corruption following power-loss.
• Bug fix: Change the VDBE so that all registers are initialized to Invalid instead of NULL.
• Bug fix: Fix problems that can result from 32-bit integer overflow.

• The FTS4 full-text search engine has been enhanced such that tokens in the search string (on the right-hand side of the MATCH operator) that begin with “^” must be the first token in their respective columns in order to match. Potentially Incompatible Change: Formerly, “^” characters in the search string were simply ignored. Hence, if a legacy application was including “^” characters in FTS4 search strings, thinking that they would always be ignored, then those legacy applications might break with this update. The fix is simply remove the “^” characters from the search string.
• Added options SQLITE_DBSTATUS_CACHE_HIT and SQLITE_DBSTATUS_CACHE_MISS to the sqlite3_db_status interface.
• Improved ANALYZE now creates the sqlite_stat3 table, a much more capbable enhancement of the previous sqlite_stat2.
• Enhance the sqlite3_data_count interface so that it can be used to determine if SQLITE_DONE has been seen on the prepared statement.
• Added the SQLITE_FCNTL_OVERWRITE file-control by which the SQLite core indicates to the VFS that the current transaction will overwrite the entire database file.
• Increase the default lookaside memory allocator allocation size from 100 to 128 bytes.
• Enhanced the query planner so that it can factor terms in and out of OR expressions in the WHERE clause in an effort to find better indices.
• Overflow pages can now be read directly from the database file, bypassing the page cache. This might improve read performance of large BLOBs (DISQLite3 Pro only).
• Remove limits on the magnitude of precision and width value in the format specifiers of the sqlite3_mprintf family of string rendering routines.
• Fix a bug that prevent ALTER TABLE | ALTER TABLE … RENAME from working on some virtual tables in a database with a UTF16 encoding.
• Fix a bug in ASCII-to-float conversion that causes slow performance and incorrect results when converting numbers with ridiculously large exponents.
• Fix a bug that causes incorrect results in aggregate queries that use multiple aggregate functions whose arguments contain complicated expressions that differ only in the case of string literals contained within those expressions.
• Fix a bug that prevented the page_count and quick_check pragmas from working correctly if their names were capitalized.
• Fix a bug that caused VACUUM to fail if the count_changes pragma was engaged.
• Fix a bug in virtual table implementation that causes a crash if an FTS4 table is dropped inside a transaction and a SAVEPOINT occurs afterwards.
• New “external merge sort” algorithm used to implement ORDER BY and GROUP BY and also to presort the content of an index for CREATE INDEX (DISQLite3 Pro only). The new algorithm does approximately the same number of comparisons and I/Os as before, but the I/Os are much more sequential and so runtimes are greatly reduced when the size of the set being sorted is larger than the filesystem cache. The performance improvement can be dramatic – orders of magnitude faster for large CREATE INDEX commands. On the other hand, the code is slightly slower (1% or 2%) for a small CREATE INDEX. Since CREATE INDEX is not an operation that commonly occurs on a speed-critical path, we feel that this tradeoff is a good one. The slight slowdown for small CREATE INDEX statements might be recovered in a future release. ORDER BY and GROUP BY operations should now be faster for all cases, large and small.
• Improved the windows VFS to better defend against interference from anti-virus software.
• Improved query plan optimization when the DISTINCT keyword is present on SELECT statements.
• Increase the default size of a lookahead cache line from 100 to 128 bytes.
• Bug fix: Virtual tables now handle IS NOT NULL constraints correctly.
• Bug fixes: Correctly handle nested correlated subqueries used with indices in a WHERE clause.

• Support Delphi XE2 Win64.
• SQLite's ATTACH stopped to recognize the DISQLite3 encryption codec so no password was set for attached databases. Fixed.
• Fix buffer allocation in Demos\DISQLite3_Full_Text_Search\DISQLite3Tokenizer.pas. Long tokens were not recognized at full length.
• Add column sorting to the DISQLite3_Full_Text_Search demo.

• Support Delphi XE2 Win32.
• Add DISQLite3_FastReport demo project.

• Add support for URI filenames. URI filenames are disabled by default (for backwards compatibility) but applications are encouraged to enable them since incompatibilities are likely to be exceedingly rare and the feature is useful.
• Add the sqlite3_vtab_config interface in support of ON CONFLICT clauses with virtual tables.
• The virtual table interface has been enhanced to support SAVEPOINT and ON CONFLICT clause processing, and the built-in RTREE and FTS3/FTS4 have been augmented to take advantage of the new capability. This means, for example, that it is now possible to use the REPLACE command on FTS3/FTS4 and RTREE tables.
• Avoid unnecessary reparsing of the database schema.
• Added support for the FTS4 prefix option and the FTS4 order option.
• Allow WAL-mode databases to be opened read-only as long as there is an existing read/write connection.
• Fix a problem with WAL mode which could cause transactions to silently rollback if the cache_size is set very small (less than 10) and DISQLite3 comes under memory pressure.

• Integer arithmetic operations that would have resulted in overflow are now performed using floating-point instead.
• New “fuzzer” Virtual Table (DISQLite3 Pro only).
• New “wholenumber” Virtual Table.
• Increased the version number on the VFS object to 3 and added new methods xSetSysCall, xGetSysCall, and xNextSysCall used for testing.
• Increase the maximum number of attached databases from 30 to 62, even though the default value remains at 10.
• Enhance the ANALYZE command to support the name of an index as its argument, in order to analyze just that one index.

This is an intermediate release based on an a pre SQLite 3.7.6 snapshot. Recommended only to make use of the newly added functionality.

• Enable support for loadable extensions. Extension modules may be build using DISQLite3Ext.
• Added the sqlite3_wal_checkpoint_v2 interface and enhanced the wal_checkpoint pragma to support blocking checkpoints.
• Improvements to the query planner so that it makes better estimates of plan costs and hence does a better job of choosing the right plan.
• Fix a bug which prevented deferred foreign key constraints from being enforced when sqlite3_finalize was not called by one statement with a failed foreign key constraint prior to another statement with foreign key constraints running.
• Enhancements to FTS4: Add the compress and uncompres options and direct access to the full-text index.
• Bug fixes.

• Added the SQLITE_DBSTATUS_LOOKASIDE_HIT, SQLITE_DBSTATUS_LOOKASIDE_MISS_SIZE, and SQLITE_DBSTATUS_LOOKASIDE_MISS_FULL options for the sqlite3_db_status interface.
• Updates to sqlite3_stmt_readonly so that its result is well-defined for all prepared statements and so that it works with VACUUM.
• Fix a bug involving frequent changes in and out of WAL mode and VACUUM that could (in theory) cause database corruption.
• Enhance the sqlite3_trace mechanism so that nested SQL statements such as might be generated by virtual tables are shown but are shown in comments and without parameter expansion. This greatly improves tracing output when using the FTS3/4 and/or RTREE virtual tables.
• Change the xFileControl() methods on all built-in VFSes to return SQLITE_NOTFOUND instead of SQLITE_ERROR for an unrecognized operation code.
• The SQLite core invokes the SQLITE_FCNTL_SYNC_OMITTED file control to the VFS in place of a call to xSync if the database has PRAGMA synchronous set to OFF.
• Testing shows that this release is about 1% or 2% faster than the previous release for most operations.

• Added the sqlite3_blob_reopen interface to allow an existing sqlite3_blob object to be rebound to a new row. Also use this to improve the performance of FTS3.
• Added the sqlite3_create_function_v2 interface that includes a destructor callback.
• VFSes that do not support shared memory are allowed to access WAL databases if PRAGMA locking_mode is set to EXCLUSIVE.
• Added support for custom r-tree queries using application-supplied callback routines to define the boundary of the query region.
• Enhancements to EXPLAIN QUERY PLAN.
• Added the sqlite3_stmt_readonly interface.
• Added PRAGMA checkpoint_fullfsync.
• Added the SQLITE_FCNTL_FILE_POINTER option to sqlite3_file_control.
• Added support for FTS4 and enhancements to the FTS matchinfo function.
• Added the sqlite3_soft_heap_limit64 interface as a replacement for sqlite3_soft_heap_limit.
• The default page cache strives more diligently to avoid using memory beyond what is allocated to it by SQLITE_CONFIG_PAGECACHE. Or if using page cache is allocating from the heap, it strives to avoid going over the sqlite3_soft_heap_limit64, even if SQLITE_ENABLE_MEMORY_MANAGEMENT is not defined at compile time.
• The ANALYZE command now gathers statistics on tables even if they have no indices.
• Tweaks to the query planner to help it do a better job of finding the most efficient query plan for each query.
• Enhanced the internal text-to-numeric conversion routines so that they work with UTF8 or UTF16, thereby avoiding some UTF16-to-UTF8 text conversions.
• Fix a problem that was causing excess memory usage with large WAL transactions in win32 systems.
• The interface between the VDBE and B-Tree layer is enhanced such that the VDBE provides hints to the B-Tree layer letting the B-Tree layer know when it is safe to use hashing instead of B-Trees for transient tables.
• A very obscure bug associated with the or optimization was fixed.
• Miscellaneous documentation enhancements.

• Delphi XE support.
• Add thread safety to help index and update FAQ.

• Fix an old and very obscure bug that can lead to corruption of the database free-page list when incremental_vacuum is used multiple times to partially reduce the size of a database file that contains many hundreds of unused database pages. The bug is impossible to hit without using incremental_vacuum and is very difficult to hit even with incremental_vacuum. And the kind of corruption that the bug causes can usually be fixed simply by running VACUUM.
• Added new commands SQLITE_DBSTATUS_SCHEMA_USED and SQLITE_DBSTATUS_STMT_USED to the sqlite3_db_status interface, in order to report out the amount of memory used to hold the schema and prepared statements of a connection.
• Increase the maximum size of a database pages from 32KiB to 64KiB. A database with 64KiB pages will not be readable or writable by older versions of SQLite. Note that further increases in page size are not feasible since the file format uses 16-bit offsets to structures within each page.
• Use the LIKE optimization even if the right-hand side string contains no wildcards.
• Added the SQLITE_FCNTL_CHUNK_SIZE verb to the sqlite3_file_control interface, to cause database files to grow in large chunks in order to reduce disk fragmentation.
• Fixed a bug in the query planner that caused performance regresssions relative to SQLite 3.6.23.1 on some complex joins.
• Performance improvement to TDISQLite3Database.Execute when processing very large SQL scripts.

• Fix a potential database corruption problem that can result if the same database file is alternately written by DISQLite3 version 2.3.0 (SQLite version 3.7.0) and DISQLite3 2.2.2 (SQLite 3.6.23.1) or earlier.
• Fix a possible performance regression caused by the introduction of automatic indexing.

The differences between DISQLite3 2.3.0.1 and DISQLite3 2.3.0 are minimal, but because of the possibility of database corruption, upgrading is highly recommended.

• Added support for write-ahead logging (DISQLite3 Pro only).
• Query planner enhancement – automatic transient indices are created when doing so reduces the estimated query time.
• Query planner enhancement – the ORDER BY becomes a no-op if the query also contains a GROUP BY clause that forces the correct output order.
• Add the SQLITE_DBSTATUS_CACHE_USED verb for sqlite3_db_status.
• The logical database size is now stored in the database header so that bytes can be appended to the end of the database file without corrupting it.

• Fix a missing “sync” that when omitted could lead to database corruption if a power failure or OS crash occurred just as a ROLLBACK operation was finishing.
• Fix that INSERTing into an RTREE returns the correct sqlite3_last_insert_rowid.
• Correctly handle strings with zero tokens in the FTS3 offsets() function.
• Enhance FTS3 to take advantage of the MAX() optimization.
• Improved error message when the right-hand operand of MATCH in FTS3 is not a valid search expression.
• Fix crash compiling “INSERT INTO tbl DEFAULT VALUES” if tbl has BEFORE or INSTEAD OF triggers.
• Fix an out of memory error on ANALYZE with STAT2 enabled with zero-length data.

• Added the secure_delete pragma.
• Added the sqlite3_compileoption_used and sqlite3_compileoption_get interfaces as well as the compile_options pragma and the sqlite_compileoption_used() and sqlite_compileoption_get() SQL functions.
• Added the sqlite3_log interface together with the SQLITE_CONFIG_LOG verb to sqlite3_config.
• Improvements to FTS3.
• The integrity_check pragma is enhanced to detect out-of-order rowids.
• Various minor bug fixes and performance enhancements.
• Completely revised the help generator to ease navigation and improve readability. Send your comments!

IMPORTANT INCOMPATIBILITY CHANGE – please read carefully:

Up to this version, the sqlite3 type was an untyped Pointer. This worked fine when used correctly, but caused hard to identify errors when not. Incompatible variable types are usually detected by the compiler's type checking. But as a matter of fact, Delphi does not generate type errors for untyped Pointer. Code would compile happily if sqlite3 was used instead of sqlite3_stmt, and Delphi would not even object to TDISQLite3Database in place of sqlite3.

Instead, problems surfaced at runtime. They often materialized as access violations with no meaningful error message and could drive helpless developers into despair. When contacting support, some of them even attributed the problem to a bug in DISQLite3. Even though this was not correct, they certainly had a point. So after careful consideration I decided to change all untyped Pointer types to pointers to record structures.

For record pointers, Delphi can detect type conflicts at compile time, so the likelihood of critical runtime errors is greatly reduced. For most stringent type checking, I recommend to compile with “Typed @ operator” always enabled. This can be set as a project option or in code using {$T+} or {$TYPEDADDRESS ON}. With typed @ operator enabled, Delphi also detects incompatible output variable types passed to sqlite3_open, sqlite3_prepare_v2, and similar routines.

Unfortunately, the modified strong typing might ask to change existing code before it compiles. Luckily, however, with {$TYPEDADDRESS ON} Delphi now catches all type incompatibilities so they can be fixed easily. For smaller projects, this is a simple search and replace (older IDEs) or code refactoring (newer ones).

Automatic conversion for larger projects is at hand here:

• https://www.yunqa.de/delphi/downloads/DISQLite3_Type_Converter.zip

DISQLite3_Type_Converter.zip contains a precompiled version of the JVCLConverter utility plus the conversion data file. It takes less than a minute to update the DISQLite3 Demos from previous versions: Run JVCLConverter.exe, select the source code folder, and convert. Converting other projects should be just as quickly.

Limit to auto-conversion: The PPointerArray type (mostly used in sqlite3_create_function callbacks) is a standard Delphi type. Converting this automatically might cause unintended side effects if used outside the DISQLite3 context. Therefore PPointerArray must be manually changed to sqlite3_value_ptr_array_ptr.

This table lists all modified, renamed, or deleted type names and their new counterparts. Only those at the top of the list should be in frequent use:

Bugs fixed in this version:

• Problems that might cause incorrect query results:
  • Queries that have a WHERE clause of the form (x AND y) OR z where x and z come from one table of a join and y comes from a different table.
  • Queries where the use of the CAST operator in the WHERE clause could lead to incorrect results if the column being cast to a new datatype is also used in the same WHERE clause without being cast.
  • Expressions with different COLLATIONs could compare equal and lead to wrong query results.
• When reusing an instance of TDISQLite3Database, the internal transaction counter was not reset when opening another database so its InTransaction property could return wrong results. Fixed.
• Other minor bug fixes and performance enhancements, especially for the FTS3 full text search.

• The SQL output resulting from sqlite3_trace is now modified to include the values of bound parameters.
• Optimizer enhancement: prepared statements are automatically re-compiled when a binding on the right hand side of a LIKE operator changes or when any range constraint changes for analyzed tables.
• Performance optimizations (12% reduction in the number of CPU operations) targetting specific use cases. Changes include the ifnull(), coalesce(), and substr() SQL functions plus various code refactorizations.
• More flexible Unicode tokenizer for the DISQLite3_Full_Text_Search demo. Added option for word stemming in 15 languages using YuStemmer.
• New FTS3 documentation is now available.
• Various minor bug fixes and documentation enhancements.
• New DISQLite3_Backup demo is a fully functional console application to back up any “hot” SQLite database.

• Native support for foreign key constraints. Foreign key constraints are disabled by default. Use the foreign_keys pragma to turn them on.
• Generalized the IS and IS NOT operators to take arbitrary expressions on their right-hand side.
• Recursive triggers support. The older non-recursive behavior of triggers is still the default behavior. Recursive triggers are activated using the recursive_triggers pragma. In addition to allowing triggers to call themselves (either directly or indirectly) the new capability also fires DELETE triggers on rows that are removed from a table as a result of REPLACE conflict resolution processing. Non-recursive triggers are still the default behavior since this is least likely to cause problems for existing applications. However, it is anticipated that triggers will become recursive by default sometime in the future. At that point, applications that want to continue using the older non- recursive trigger behavior will need to use the recursive_triggers pragma to disable recursive triggers.
• Refinements to the query planner that help to choose better plans for joins where in the past it was selecting suboptimal query plans. Running the ANALYZE command will collect histogram data on indices. The use of histograms improves the query planning performance even more.
• Added the SQLITE_OPEN_SHAREDCACHE and SQLITE_OPEN_PRIVATECACHE flags for sqlite3_open_v2 used to override the global shared cache mode settings for individual database connections.
• Added improved version identification features: New sqlite3_sourceid interface and SQL function sqlite_source_id().
• Obscure bug fix on triggers, plus other bux fixes.

• Delphi 2010 support.
• Fix a threading issue with TDISQLite3Database by wrapping all calls to its FStatements field with mutexes.
• Fix sqlite3_key so that it successfully sets a correct password after an incorrect one.

• Mark sqlite3_prepare and sqlite3_prepare16 as deprecated. They are retained for backwards compatibility, but their use is discounraged. Instead, sqlite3_prepare_v2 and sqlite3_prepare16_v2 should be used, which work much better with SQLITE_SCHEMA errors.
• Expose the sqlite3_strnicmp interface for use by extensions and applications.
• Remove the restriction on virtual tables and shared cache mode. Virtual tables and shared cache can now be used at the same time.
• Modify FTS3 so that in terms like 'column_name:token' the column_name is interpreted in a case-insenstive fashion.
• Fix a double-free that can occur when using the FTS3 '-' operator.
• Other minor bug fixes.

• Fix a bug that occasionally causes INSERT or UPDATE operations to fail on an indexed table that has a self-modifying trigger.
• Internal representation of SQL expressions use less memory.
• Reduce the amount of stack space used.
• The sqlite3_create_function family of interfaces now return SQLITE_MISUSE instead of SQLITE_ERROR when passed invalid parameter combinations.
• When new tables are created using CREATE TABLE … AS SELECT … the datatype of the columns is the simplified SQLite datatype (TEXT, INT, REAL, NUMERIC, or BLOB) instead of a copy of the original datatype from the source table.
• Resolve race conditions when checking for a hot rollback journal.
• The sqlite3_shutdown interface frees all mutexes.
• Enhanced robustness against corrupt database files.
• Fix a code generator bug that can cause incorrect query results under obscure circumstances.
• Fix a bug in group_concat().
• Fix a performance bug in the pager cache.
• Fix a bug in the sqlite3_backup implementation that can lead to a corrupt backup database.
• Allow empty TDISQLite3Database.DatabaseName to create temporary databases. A warning is issued a design time.
• Other minor bug fixes and performance optimizations.

• New asynchronous backend module (DISQLite3 Pro only). It gives the illusion of faster response times by pushing slow write operations into a background thread. The tradeoff for faster response time is that more memory is required (to hold the content of the pending writes) and if a power failure or program crash occurs, some transactions that appeared to have committed might end up being rolled back upon restart.
• Performance enhancements in the btree and pager layers and in the query optimizer. Certain workloads can be as much as twice as fast as the previous release, though 10% faster is a more typical result.
• Enhanced the query optimizer so that queries against virtual tables that contain OR and IN operators in the WHERE clause are now able to use indexes.
• Many minor bug fixes.

• Fix a bug that causesd a segfault when running a count(*) on the sqlite_schema table of an empty database.
• Fix a bug that causesd a segfault that when inserting into a table using a DEFAULT value where there is a function as part of the DEFAULT value expression.
• Other minor bug fixes.

• Fixed a bug that caused database corruption when an incremental_vacuum is rolled back in an in-memory database. If an incremental_vacuum was rolled back in an in-memory database, the database would often go corrupt. This only happened for in-memory databases. On-disk databases were unaffected. And the corruption only appeared if an incremental vacuum was rolled back. Upgrading is recommended for all applications, especially those that make use of in-memory databases and/or incremental vacuum.
• Fixed a bug in the lookaside memory allocator relating to shared cache mode. If a database connection closed before others, it could prematurely deallocate critical elements of the shared schema.
• Added the sqlite3_unlock_notify interface (DISQLite3 Pro only).
• Added the reverse_unordered_selects pragma (DISQLite3 Pro only).
• The default page size is automatically adjusted to match the capabilities of the underlying filesystem.
• Performance improvements for “count(*)” queries (DISQLite3 Pro only).
• Reduce the amount of heap memory used, especially by TRIGGERs.

• New Hot Backup Interface (DISQLite3 Pro only).
• Improve TDISQLite3UniDirQuery Provider Support when inserting and updating BLOB fields with #0 values and NULL BLOBs.
• For PRIMATY KEY columns, Use the ftAutoInc field type and exclude pfInUpdate from ProviderFlags in TDISQLite3UniDirQuery. This helps to handle multiple insert problems with TClientDataSet.
• Update DISQLite3_ClientDataSet_Grid.dpr demo to use decreasing negative RowIDs so multiple inserts do not conflict.
• Various minor bug fixes.
• Type changes to sqlite3_context and sqlite3_context_ptr. This an unfortunate incompatibility change. To find out if you are affected, recompile your code and watch for compiler errors. If so, update your code by replacing “sqlite3_context” with “sqlite3_context_ptr”.
• Improvements to the syntax bubble diagrams.

• Added support for nested transactions.
• Enhanced the query optimizer so that it is able to use multiple indices to efficiently process OR-connected constraints in a WHERE clause.
• Added support for parentheses in FTS3 query patterns using the SQLITE_ENABLE_FTS3_PARENTHESIS compile-time option (source code only).
• Fix a cache coherency problem that could lead to database corruption.
• Fix two bugs, which when combined might result in incorrect query results. Both bugs were harmless by themselves; only when they team up do they cause problems.

• Add TDISQLite3UniDirQuery support for TClientDataSet.RefreshRecord.
• Fixes for various minor bugs.

• Fix a bug in the b-tree delete algorithm that seems like it might be able to cause database corruption.
• Fix a memory leak that can occur following a disk I/O error.
• Fix a bug in the page cache that can lead database corruption following a rollback.
• Add the application-defined page cache.
• TDISQLite3UniDirQuery now sets the TField.Required property for columns declared as NOT NULL.
• Add the TDISQLite3DataSetImporter.OnImportError event to catch and log exceptions (which may be caused by, for example, invalid data conversions on certain records) and continue the import.
• Also in TDISQLite3DataSetImporter, call the progress event for every single record instead of every 16th one only.
• Fix a memory overrun in TDISQLite3DataSetImporter in the line wrap functions for textual blob data.
• Other minor bug fixes and optimizations.

• Delphi 2009 support.
• Add the MEMORY option to PRAGMA journal_mode.
• Added the sqlite3_db_mutex interface.
• Added the sqlite3_extended_errcode interface.
• The COMMIT command now succeeds even if there are pending queries. It returns SQLITE_BUSY if there are pending incremental BLOB I/O requests.
• The error code is changed to SQLITE_BUSY (instead of SQLITE_ERROR) when an attempt is made to ROLLBACK while one or more queries are still pending.
• Add option support for LIMIT and ORDER BY clauses on DELETE and UPDATE statements (DISQLite3 Professional only).
• Added the sqlite3_stmt_status interface for performance monitoring.
• Add the INDEXED BY clause.
• Added the TRUNCATE option to PRAGMA journal_mode
• Performance enhancements to tree balancing logic in the B-Tree layer.
• The SQL language documentation is converted to use syntax diagrams instead of BNF.
• Bug fixes.

• Added the lookaside memory allocator for a speed improvement in excess of 15% on some workloads. (Your mileage may vary.)
• Added the SQLITE_CONFIG_LOOKASIDE verb to sqlite3_config to control the default lookaside configuration.
• Added verbs SQLITE_STATUS_PAGECACHE_SIZE and SQLITE_STATUS_SCRATCH_SIZE to the sqlite3_status interface.
• Modified SQLITE_CONFIG_PAGECACHE and SQLITE_CONFIG_SCRATCH to remove the “+4” magic number in the buffer size computation.
• Added the sqlite3_db_config and sqlite3_db_status interfaces for controlling and monitoring the lookaside allocator separately on each database connection.
• Numerious other performance enhancements.
• Miscellaneous minor bug fixes.