Parser Improvements & Fixes + Error Docs Hyperlinks

Added

• parser: validation for default usages (#1175)

  -- ok
  insert into t values (default);

  -- err
  select default;

• linter: setup hyperlinks for rule names in terminal (#1174)

  If your terminal supports it, then clicking on the rule name will navigate
  you to the docs.

Fixed

• fmt: upgrade tiny_pretty to latest (#1173)

  We no longer wrap statements unless necessary.

• parser: fix parsing period column (#1172)

  previously we'd error on:

  create table t (
    symbol text,
    period text,
    year int,
    quarter int,
    primary key (symbol, period, year, quarter)
  );

More Parser Validation + CI Hardening

v2.54.0 - 2026-05-24

Added

• lexer: add parsing and erroring for num trailing suffix (#1159)

  error[syntax-error]: trailing junk after numeric literal
    ╭▸ stdin:1:11
    │
  1 │ SELECT 0x0y;
    ╰╴          ━
  

• parser: warn about empty / invalid params & empty quoted idents (#1164)

  error[syntax-error]: empty delimited identifier
    ╭▸ stdin:1:8
    │
  1 │ select "", $, $2147483648
    ╰╴       ━━
  error[syntax-error]: missing parameter number
    ╭▸ stdin:1:12
    │
  1 │ select "", $, $2147483648
    ╰╴           ━
  error[syntax-error]: parameter number too large
    ╭▸ stdin:1:15
    │
  1 │ select "", $, $2147483648
    ╰╴              ━━━━━━━━━━━
  

• parser: warn about invalid octal/hex/binary digits (#1163)

  error[syntax-error]: invalid digit for a base 2 literal
    ╭▸ stdin:1:12
    │
  1 │ select 0b104, 0o7719, 0xg
    ╰╴           ━
  error[syntax-error]: invalid digit for a base 8 literal
    ╭▸ stdin:1:20
    │
  1 │ select 0b104, 0o7719, 0xg
    ╰╴                   ━
  error[syntax-error]: trailing junk after numeric literal
    ╭▸ stdin:1:25
    │
  1 │ select 0b104, 0o7719, 0xg
    ╰╴                        ━
  

• parser: improve error reporting for malformed literals (#1162)

  error[syntax-error]: trailing junk after positional parameter
    ╭▸ stdin:1:10
    │
  1 │ SELECT $1a;
    ╰╴         ━
  

  instead of

  error[syntax-error]: trailing junk after positional parameter
    ╭▸ stdin:1:10
    │
  1 │ SELECT $1a;
    ╰╴       ━━━
  

• parser: improve lexing numbers (#1161)

  select .4;

  now produces:

  NUMERIC_NUMBER@7..9 ".4"
  

  instead of:

  INT_NUMBER@7..9 ".4"
  

• ide: goto def for t.c%type (#1161)

  create table t(a int, b text);
  --           ^ dest
  create function f(x t.a%type) returns s.t.b%type
  --                  ^ source
    as $$ select 'hello'::text $$ language sql;

• ide: find refs for types like bit (#1153)

  create type pg_catalog.bit;
  --                     ^^^ source
  
  create function pg_catalog.bit(bigint, integer) returns bit
  --                                                      ^^^ ref
    language internal;

• ide: add hover for string literals (#1155)

  Now we decode the escape sequences for string literals on hover and show the
  value up to the first new line.

• ide: improve numeric literal type inference (#1156)

  select 2147483647; -- type: integer
  select 2147483648; -- type: bigint
  select 100000000000000000000000; -- type: numeric

• fmt: literals & binary operators (#1169)

  Format binary operators and literals.

  -- before
  select TRUE  and  FALSE;
  select X'AF';
  
  -- after
  select true and false;
  select
    x'AF';

• fmt: unquote column aliases when possible (#1168)

  -- before
  select 1 as "foo";
  
  -- after
  select 1 as foo;

Changed

• harden github action workflows (#1165). Thanks @chdsbd!

Fixed

• lexer: fix unicode escape string issue (#1158)
• vscode: update TextMate grammar to support other numeric literal kinds (#1157)

v2.53.0 - 2026-05-17

Added

• parser: validate string continuations (#1137)
• parser: improve create rule ast (#1147)

Changed

• parser: update ast nodes to include their trailing semicolons (#1145, #1148)

Fixed

• install: fix windows npx install bug (#1140)

Fix NPX Windows Install + Parser Improvements

Added

• parser: validate string continuations (#1137)
• parser: improve create rule ast (#1147)

Changed

• parser: update ast nodes to include their trailing semicolons (#1145, #1148)

Fixed

• install: fix windows npx install bug (#1140)

Fix NPM Package Publishing

Changed

Attempt to fix npm install method

Improved Parser Validation for Strings + Updated Install Method

Added

• parser: validation for bit, byte, and escape string types (#1132)

  select b'01' '10';
  select x'0F' '10';
  select e'foo' 'bar';

  now gives:

  error[syntax-error]: Expected new line or comma between string literals
    ╭▸ stdin:1:13
    │
  1 │ select b'01' '10';
    ╰╴            ━
  error[syntax-error]: Expected new line or comma between string literals
    ╭▸ stdin:2:13
    │
  2 │ select x'0F' '10';
    ╰╴            ━
  error[syntax-error]: Expected new line or comma between string literals
    ╭▸ stdin:3:14
    │
  3 │ select e'foo' 'bar';
    ╰╴             ━

• parser: validation for escape sequences (#1125, #1123, #1122, #1128, #1129)

  select U&'wrong: !061' UESCAPE '!';
  select U&"wrong: \06" UESCAPE '\';

  now gives:

  error[syntax-error]: Unicode escape requires 4 hex digits: !XXXX
    ╭▸ stdin:1:20
    │
  1 │   select U&'wrong: !061' UESCAPE '!';
    ╰╴                   ━━━━
  error[syntax-error]: Unicode escape requires 4 hex digits: \XXXX
    ╭▸ stdin:2:20
    │
  2 │   select U&"wrong: \06" UESCAPE '\';
    ╰╴                   ━━━

Changed

• ci: update npm based install method (#1133)

  Instead of fetching the binary via an install script we use an optional
  peer dependency, mirroring ESBuild and Sentry.

Fixed

• cli: fix missing binary exit code (#1130)

  Before if the install script didn't run the NPM JS shim would exit without
  erroring. Thanks @nwalters512!

• parser: parsing unicode escape idents in cast position (#1127)

  select 2::U&"!0069!006E!0074!0038" UESCAPE '!' from t;

  now parses without error

File Level Disable Comment For `assume_in_transaction`

v2.51.0 - 2026-05-06

Added

• linter: add file-level override to disable assume_in_transaction (#996). Thanks @reteps!

  When using assume-in-transaction, you can disable it on a per file basis using a comment:

  -- squawk-disable-assume-in-transaction

• syntax: add validation for select into (#1116)

  Warn about invalid usages, such as:

  • select 4 a union select 5 a into t;
  • with t as (select 1 a into t) select * from t;
  • select * from t where a in (select 1 a into t);

  and more!

• parser: add empty statement to ast (#1111)

  ;
  ;
  ;

  now gives:

  SOURCE_FILE@0..5
    EMPTY_STMT@0..1
      SEMICOLON@0..1 ";"
    WHITESPACE@1..2 "\n"
    EMPTY_STMT@2..3
      SEMICOLON@2..3 ";"
    WHITESPACE@3..4 "\n"
    EMPTY_STMT@4..5
      SEMICOLON@4..5 ";"
  

• ide: code actions create table as <-> select into + parser fix (#1113)

  Create table and select into are roughly interchangable, so this code action
  supports switching between.

• ide: show comments for types and columns (#1110)

• ide: hover infer more types for columns (#1109, #1108)

Changed

• parser: cleanup insert & create rule ast grammar to be more strict (#1112)

Fixed

• vscode: fix text mate grammar for multiple stmts on a line (#1115)
• ide: fix infer column name w/ column aliases and goto def (#1118)
• ide: fix output column alias resolution in group by / order by (#1117)
• ide: fix inherit/like resolution + cte fixes (#1107)

New Rules: Prefer Repack, Identifier Too Long, Require Concurrent Reindex, Require Concurrent Partition Detach + A Lexer Fix

Added

• linter: new rule: prefer-repack (#1105)

  repack isn't out yet, coming in PG19, but when it's out:

  -- instead of
  cluster foo;
  -- or
  vacuum full foo;
  -- use
  repack (concurrently) foo;

• linter: new rule: require-concurrent-reindex (#1104)

  -- instead of
  reindex table foo;
  -- use
  reindex table concurrently foo;

• linter: new rule: require-concurrent-partition-detach (#1103)

  -- instead of
  alter table t detach partition p;
  -- use
  alter table t detach partition p concurrently;

• linter: new rule: identifier-too-long (#1102)

  Postgres truncates identifiers that are too long, we now warn about this.

  create table table_very_long_very_long_very_long_very_long_very_long_very_long (c bigint);

  warning[identifier-too-long]: `table_very_long_very_long_very_long_very_long_very_long_very_long` is too long and will be truncated to 63 bytes.
    ╭▸ stdin:1:14
    │
  1 │ create table table_very_long_very_long_very_long_very_long_very_long_very_long (c bigint);
    │              ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
    ╭╴
  1 - create table table_very_long_very_long_very_long_very_long_very_long_very_long (c bigint);
  1 + create table table_very_long_very_long_very_long_very_long_very_long_very_lo (c bigint);
    ╰╴
  

  Thanks @subnix for coming up with this idea!

• linter: parallel file checking (#1099)

  We now check files in parallel using rayon, which gives a nice speed up of
  22-55%.

• lexer/ide: fix lexing of non-ascii + fix case folding in goto def (#1101)

  Previously we case folding all of the characters in our unquoted identifiers
  including unicode. This differed from Postgres behavior which only folded ascii.

  Now we correctly error instead of resolving in the following:

  with t as (select 1 Äpfel)
  select äpfel from t;
  --     ^ goto def no longer resolves

  Additionally fixed a bug in the lexer where we weren't lexing unicode identifiers.

  For example, the following now lexes:

  with t as (select 1 🦀)
  select 🦀 from t;

Spring Cleaning + Some IDE Improvements

Added

• ide: goto def for select into (#1086)

  select 1 a into t;
  --       ^ 2. dest
  select a from t;
  --     ^ 1. goto def src

  and

  select 1 a into t;
  --              ^ 2. dest
  select a from t;
  --            ^ 1. goto def src

• ide: goto def with view & qualified column (#1081)

  create view v as select 1 id, 2 b;
  --          ^ 2. dest
  select v.id from v;
  --     ^ 1. goto def src

• ide: goto def from graph_table to create property graph (#1077)

  create property graph myshop vertex tables (t key (a) no properties);
  --                    ^ 2. dest
  select 1 from graph_table (myshop
  --                         ^ 1. goto def src
    match (n is t)
    columns (1 as x));

• fmt: more work towards full select and create table support (#1079)

  -- before
  create table t(a int,b text);
  
  -- after
  create table t(
    a int,
    b text
  );

  -- before
  select array[[1,2],[3,4]];
  
  -- after
  select array[[1, 2], [3, 4]];

Fixed

• ide: fix shadowing cte column w/ * hover + func call syntax for cte & views (#1094)

  -- shadowing
  create table t(a int, b int);
  with
    t as (
      select 1
    ),
    -- yy overrides y since there's only 1 column in the *
    u(x, yy) as (
      select *, 2 y, 3 z from t
    )
  select y from u;
  --     ^ 1. goto def src - doesn't resolve because of override above

  -- function call syntax for cte column
  with cte as (select 1 as a)
  --                       ^ 2. dest
  select a(cte) from cte;
  --     ^ 1. goto def src

  -- function call syntax for view column
  create view v as select 1 as a;
  --                           ^ 2. dest
  select a(v) from v;
  --     ^ 1. goto def src

Internal

• ide/server/wasm: use salsa for binder caching (#1078)
• ide: cleanup goto def & find ref tests to prep for extension support (#1080)
• ide: split resolve into resolve/collect/ast_nav & dedupe code (#1085)
• ide: consolidate name helpers (#1090)
• ide: update return types in resolve_name_ref to use Location (#1089)
• ide: simplify hovering for names (#1088)
• ide: refactor hover & find refs to use goto_def (#1087)
• ide: refactor resolve and fix a couple edge cases (#1094)
• ide: refactor resolve to use name related functions (#1092)
• ide: refactor code actions into their own modules (#1091)

New Rule: require-table-schema + IDE & Parser Improvements

Added

• linter: add require-table-schema rule (#1046, #1064, #1073). Thanks @Flaiers!

  Note: this rule is disabled by default and must be enabled via the new
  --include flag.

  -- error
  create table posts(id bigint);
  
  -- okay
  create table public.posts(id bigint);

• parser: improve error recovery for misplaced join clauses (#1065)

  The following now gives a concise error message:

  -- join after the where
  select * from t
  where x > 1
  join k on true;

• parser: support more statement kinds in create schema parsing (#1061)

• ide: semantic syntax highlighting improvements (#1059, #1060, #1068)

  We now highlight all the tokens in types. We were missing some like setof.

  Additionally, we highlight names by their underlying kind i.e., table, function, column.
  This is powered by goto def.

  You can't tell from GitHub's highlighting, but if you use the language
  server, we now highlight the b in t.b as a function:

  create table t(a int);
  create function b(t) returns int as 'select 1' language sql;
  select b(t), t.b from t;

• ide: show function comment on hover (#1070)

• ide: code action to rewrite between != and <> (#1066)

• ide: goto def for create property graph (#1074)

Fixed

• parser: fix param parsing (#1068)

  Before, double precision was parsed as a param named double with type precision.
  Now it's parsed correctly as an unnamed param of type double precision.

  create function f(double precision) returns int8
    as 'select $1'
    language sql;

Semantic Syntax Highlighting

Added

• ide: semantic syntax highlighting (#1057, #1058, #1052)

• ide: semantic syntax highlighting (#1048)

• playground: update syntax highlighting language with codegen'd keywords (#1055, #1054)

• vscode: update syntax highlighting to support more operators (#1047)

• ide: goto def for user, current_user, session_user, & current_schema (#1056)