- Breaking/Improvement: replace_placeholders now uses exp.convert instead of exp.to_identifier.
- New: Added support for the HLL aggregate function (Snowflake, Redshift).
- New: Added support for MATCH ... AGAINST ... expression (MySQL).
- New: Added TPC-DS in the test suite.
- New: Added support for BigQuery's SELECT AS [STRUCT|VALUE] expression.
- Improvement: Allow parsing `[]` for [nested types](https://github.com/tobymao/sqlglot/commit/709421aab7ef427619eb5feb95988a0f36792b84).
- Improvement: [Normalize](https://github.com/tobymao/sqlglot/commit/58fdbf05f5e7627936ebf2036410c60396d6da72) keyword texts in the tokenizer.
- Improvement: Parse DECODE function into CASE expression when it has >2 args.
- Improvement: [Don't expand aliases in inner scopes](https://github.com/tobymao/sqlglot/commit/9a574b0e79e7532fd49929cf86396d4485091118).
- Improvement: [Allow HAVING to alias selects](https://github.com/tobymao/sqlglot/commit/2b3eb31ca2ecb81a70bb462ff3987f0749a32d0e).
- Improvement: Added support for the limit argument in Spark's SPLIT function.
- Improvement: [Fix](https://github.com/tobymao/sqlglot/commit/68aea4ab95586723ed0c3d3ab0281577cc71e40c) related to expanding forward references in selects.
- Improvement: Generate ->> for JSON_EXTRACT_SCALAR (MySQL).
- Improvement: [Fix](https://github.com/tobymao/sqlglot/commit/68aea4ab95586723ed0c3d3ab0281577cc71e40c) related to expanding alias references in the WHERE clause.
- Improvement: Fixed transpilation of some char types to Bigquery.
- Improvement: [Map MAX to GREATEST conditionally for some dialects](https://github.com/tobymao/sqlglot/commit/27483e9f3b7f3043ba1b325d58aa864db54255f1).
- Improvement: Make normalize default to CNF.
- Improvement: Fixed star expansion when USING clause is present.
- Improvement: Various performance improvements in the optimizer.
- Improvement: Various simplifications in the executor (e.g. scan, join).
- Breaking/Improvement: Fixed transpilation of the division ("/") operator, because some dialects do integer division while others do float division. For example, "a / b" will now be transpiled to "CAST(a AS DOUBLE PRECISION) / b" for Snowflake -> Postgres, and to "CAST(a / b AS INT)" for Postgres -> Snowflake.
- New: Added support for the COMMENT ON statement.
- New: Added support for old (+) JOIN syntax (Oracle).
- New: Added "root" helper method on Expression.
- New: Added "meta" property on Expression.
- Improvement: Refactored parsing of ALTER TABLE so it's easily extensible.
- Improvement: Merging subquery to out of scope join bug fixed.
- Improvement: Tokenize Jinja comments as comments.
- Improvement: Refactor POST_CREATE and POST_EXPRESSION properties.
- Improvement: Fix line numbers for strings that contain multi-line comments.
- Improvement: Fix aliasing for pivot/unpivot expressions.
- Breaking: Although this is a breaking change, it fixes and makes the behavior of escaped characters more consistent. Now SQLGlot preserves all line break inputs as they come.
- Breaking: Added python type hints in the parser module, which may result in some mypy errors.
- New: SQLGlot expressions can [now be serialized / deserialized into JSON](https://github.com/tobymao/sqlglot/commit/bac38151a8d72687247922e6898696be43ff4992).
- New: Added support for T-SQL [hints](https://github.com/tobymao/sqlglot/commit/3220ec1adb1e1130b109677d03c9be947b03f9ca) and [EOMONTH](https://github.com/tobymao/sqlglot/commit/1ac05d9265667c883b9f6db5d825a6d864c95c73).
- New: Added support for Clickhouse's parametric function syntax.
- New: Added [wider support](https://github.com/tobymao/sqlglot/commit/beb660f943b73c730f1b06fce4986e26642ee8dc) for timestr and datestr.
- New: CLI now accepts a flag [for parsing SQL from the standard input stream](https://github.com/tobymao/sqlglot/commit/f89b38ebf3e24ba951ee8b249d73bbf48685928a).
- Improvement: Fixed BigQuery transpilation for [parameterized types and unnest](https://github.com/tobymao/sqlglot/pull/924).
- Improvement: Hive / Spark identifiers can now begin with a digit.
- Improvement: Bug fixes in [date/datetime simplification](https://github.com/tobymao/sqlglot/commit/b26b8d88af14f72d90c0019ec332d268a23b078f).
- Improvement: Bug fixes in [merge_subquery](https://github.com/tobymao/sqlglot/commit/e30e21b6c572d0931bfb5873cc6ac3949c6ef5aa).
- Improvement: Schema identifiers are now [converted to lowercase](https://github.com/tobymao/sqlglot/commit/8212032968a519c199b461eba1a2618e89bf0326) unless they're quoted.
- Improvement: Identifiers with a leading underscore are now regarded as [safe](https://github.com/tobymao/sqlglot/commit/de3b0804bb7606673d0bbb989997c13744957f7c#diff-7857fedd1d1451b1b9a5b8efaa1cc292c02e7ee4f0d04d7e2f9d5bfb9565802c) and hence are not quoted.
- Breaking: types inferred from annotate_types are now DataType objects, instead of DataType.Type.
- New: the optimizer can now simplify [BETWEEN expressions expressed as explicit comparisons](https://github.com/tobymao/sqlglot/commit/e24d0317dfa644104ff21d009b790224bf84d698).
- New: the optimizer now removes redundant casts.
- New: added support for Redshift's ENCODE/DECODE.
- New: the optimizer now [treats identifiers as case-insensitive](https://github.com/tobymao/sqlglot/commit/638ed265f195219d7226f4fbae128f1805ae8988).
- New: the optimizer now [handles nested CTEs](https://github.com/tobymao/sqlglot/commit/1bdd652792889a8aaffb1c6d2c8aa1fe4a066281).
- New: the executor can now execute SELECT DISTINCT expressions.
- New: added support for Redshift's COPY and UNLOAD commands.
- New: added ability to parse LIKE in CREATE TABLE statement.
- New: the optimizer now [unnests scalar subqueries as cross joins](https://github.com/tobymao/sqlglot/commit/4373ad8518ede4ef1fda8b247b648c680a93d12d).
- Improvement: fixed Bigquery's ARRAY function parsing, so that it can now handle a SELECT expression as an argument.
- Improvement: improved Snowflake's [ARRAY and MAP constructs](https://github.com/tobymao/sqlglot/commit/0506657dba55fe71d004c81c907e23cdd2b37d82).
- Improvement: fixed transpilation between STRING_AGG and GROUP_CONCAT.
- Improvement: the INTO clause can now be parsed in SELECT expressions.
- Improvement: improve executor; it currently executes all TPC-H queries up to TPC-H 17 (inclusive).
- Improvement: DISTINCT ON is now transpiled to a SELECT expression from a subquery for Redshift.
- Breaking: [refactored](https://github.com/tobymao/sqlglot/commit/6b0da1e1a2b5d6bdf7b5b918400456422d30a1d4) the way SQL comments are handled. Before at most one comment could be attached to an expression, now multiple comments may be stored in a list.
- Breaking: [refactored](https://github.com/tobymao/sqlglot/commit/be332d10404f36b43ea6ece956a73bf451348641) the way properties are represented and parsed. The argument `this` now stores a property's attributes instead of its name.
- New: added structured ParseError properties.
- New: the executor now handles set operations.
- New: sqlglot can [now execute SQL queries](https://github.com/tobymao/sqlglot/commit/62d3496e761a4f38dfa61af793062690923dce74) using python objects.
- New: added support for the [Drill dialect](https://github.com/tobymao/sqlglot/commit/543eca314546e0bd42f97c354807b4e398ab36ec).
- New: added a `canonicalize` method which leverages existing type information for an expression to apply various transformations to it.
- New: TRIM function support for Snowflake and Bigquery.
- New: added support for SQLite primary key ordering constraints (ASC, DESC).
- New: added support for Redshift DISTKEY / SORTKEY / DISTSTYLE properties.
- New: added support for SET TRANSACTION MySQL statements.
- New: added `null`, `true`, `false` helper methods to avoid using singleton expressions.
- Improvement: allow multiple aggregations in an expression.
- Improvement: execution of left / right joins.
- Improvement: execution of aggregations without the GROUP BY clause.
- Breaking: replaced SQLGlot annotations with comments. Now comments can be preserved after transpilation, and they can appear in other places besides SELECT's expressions.
- Improvement: [refactored the schema](https://github.com/tobymao/sqlglot/pull/668) to be more lenient; before it needed to do an exact match of db.table, now it finds table if there are no ambiguities.
- Improvement: allow functions to [inherit](https://github.com/tobymao/sqlglot/pull/674) their arguments' types, so that annotating CASE, IF etc. is possible.
- Breaking : Changed AST hierarchy of exp.Table with exp.Alias. Before Tables were children's of their aliases, but in order to simplify the AST and fix some issues, Tables now have an alias property.
- New: SQLGlot now supports the [PySpark](https://github.com/tobymao/sqlglot/tree/main/sqlglot/dataframe) dataframe API. This is still relatively experimental.
- Breaking: Removed $ from tokenizer, added @ placeholders
- Improvement: Nodes can now be removed in transform and replace [8cd81c3](https://github.com/tobymao/sqlglot/commit/8cd81c36561463b9849a8e0c2d70248c5b1feb62)
- Improvement: Snowflake timestamp support
- Improvement: Property conversion for CTAS Builder
- Improvement: Tokenizers are now unique per dialect instance
