highlight(sql): Update tree-sitter-sql to 98a7fc9

[LeoniePhiline]

Changes:

Updates tree-sitter-sql from 3a3f92b29c880488a08bc2baaf1aca6432ec3380 to c5b1548ec95956c8da6393dceacf30aa349a6e9f.

Full grammar.js.

Note that to avoid further diff noise, the tree sitter JSON and C sources were moved onto the gh-pages branch.

This is why languages.toml references DerekStride/tree-sitter-sql@7cbac04 on the gh-pages branch, which contains the sources for DerekStride/tree-sitter-sql@98a7fc9 on the main branch (the latest grammar change on main).

Merged PRs:

• Create Tables/views from CTE's DerekStride/tree-sitter-sql#105
• feat: support single quotes in string literals DerekStride/tree-sitter-sql#109
• Add NOT <> expression to grammar DerekStride/tree-sitter-sql#112
• feat: add bit datatype DerekStride/tree-sitter-sql#110
• Allow keywords to be case insensitive DerekStride/tree-sitter-sql#113
• Allow INSERT tables to have aliases DerekStride/tree-sitter-sql#114
• Add additional clauses/keywords to the insert statement DerekStride/tree-sitter-sql#118
• feat: add binary and string datatypes for hive DerekStride/tree-sitter-sql#116
• Publish parser artifacts on GitHub pages build DerekStride/tree-sitter-sql#100
• Update Readme DerekStride/tree-sitter-sql#119
• feat: add binary between expression DerekStride/tree-sitter-sql#117
• feat: add more options to pg views DerekStride/tree-sitter-sql#121
• feat: add default_timestamp option DerekStride/tree-sitter-sql#123
• fix: check option as extra keyword nodes (for better highlighting) DerekStride/tree-sitter-sql#125
• Update installation instructions DerekStride/tree-sitter-sql#124
• Add support for parenthesized expressions & EXISTS subquery DerekStride/tree-sitter-sql#126
• add support for mysql enum type DerekStride/tree-sitter-sql#127
• add full outer joins DerekStride/tree-sitter-sql#130
• fix frame_definition of window functions DerekStride/tree-sitter-sql#131
• add support for mysql comment keyword in column definition DerekStride/tree-sitter-sql#132
• fix: case when with between statement DerekStride/tree-sitter-sql#134
• add optional 'unique' keyword in key_constraint DerekStride/tree-sitter-sql#136
• fix: various errors DerekStride/tree-sitter-sql#138
• allow count in window functions DerekStride/tree-sitter-sql#139
• feat: create function DerekStride/tree-sitter-sql#141
• removed count in favour of invocation DerekStride/tree-sitter-sql#143
• fix: parses parens in create table statement DerekStride/tree-sitter-sql#146
• feat: add insert overwrite from spark sql DerekStride/tree-sitter-sql#148
• feat: add mysql rename statement DerekStride/tree-sitter-sql#152
• Optimize metadata DerekStride/tree-sitter-sql#153
• fix: distinct and order by in invocation DerekStride/tree-sitter-sql#156
• ci: check tests for ERROR DerekStride/tree-sitter-sql#157

(Solved) Issue with keyword_language

During review you might notice that (keyword_language) is commented out in line 210.

It seems to me like this query should work, but as long as this keyword query is enabled, helix fails parsing with:

helix_core::syntax [ERROR] Could not parse queries for language "sql". Are your grammars out of sync? Try running 'hx --grammar fetch' and 'hx --grammar build'. This query could not be parsed: QueryError { row: 209, column: 3, offset: 3657, message: "keyword_language", kind: NodeType }

This seems to be a TSQueryError from tree-sitter's C ts_query_new.

The symbol is defined in grammar.json.

But it is not defined in node-types.json.

In parser.c, it is defined as aux_sym_keyword_language_token1. 🤔 (see also the codegen)

(keyword_language) appears to be unused and thus declassified by the tree-sitter codegen to an auxiliary symbol.

Disabling (keyword_language) in the queries list is a viable workaround. I'd completely remove it if it is removed upstream.

[the-mikedavis]

Thanks!

[the-mikedavis]

Yeah as you noted in the issue it looks like that node is unused. When the tree-sitter C library checks the queries ("query analysis") it will reject patterns that can't possibly match. When it says that for named nodes that exist in the grammar.js it either means that the named node is unused like in this case or that it's used but always aliased to something else. So if a node can't appear in the syntax tree we'll see that error.

[LeoniePhiline]

I'll add one more patch, since DerekStride/tree-sitter-sql#154 got fixed.

Yeah as you noted in the issue it looks like that node is unused. When the tree-sitter C library checks the queries ("query analysis") it will reject patterns that can't possibly match. When it says that for named nodes that exist in the grammar.js it either means that the named node is unused like in this case or that it's used but always aliased to something else. So if a node can't appear in the syntax tree we'll see that error.

Thanks!

[LeoniePhiline]

I pushed e401a61 which only fixes a COUNT(DISTINCT tbl.field) bug and adds some CI.

I updated the PR title, description and hyperlinks to reflect the addition.

The referenced commit 7cbac0472e5b8f8486ce64ffbcf1982d5cd5fc8d in languages.toml is HEAD of the gh-pages branch, which contains the generated tree-sitter C code.

[pascalkuthe]

Thanks!