https://github.com/skalt/psql_splitter

a nom parser combinator that matches a psql statement.
https://github.com/skalt/psql_splitter

parser parsing postgres postgresql psql sql

Last synced: 15 days ago
JSON representation

a nom parser combinator that matches a psql statement.

• Host: GitHub
• URL: https://github.com/skalt/psql_splitter
• Owner: SKalt
• License: bsd-3-clause
• Created: 2021-12-04T15:00:44.000Z (about 3 years ago)
• Default Branch: main
• Last Pushed: 2022-02-27T15:02:57.000Z (almost 3 years ago)
• Last Synced: 2024-10-27T07:52:24.831Z (2 months ago)
• Topics: parser, parsing, postgres, postgresql, psql, sql
• Language: Rust
• Homepage:
• Size: 20.5 KB
• Stars: 1
• Watchers: 2
• Forks: 0
• Open Issues: 0
• Metadata Files:
  • Readme: README.md
  • License: LICENSE

Awesome Lists containing this project

README

        
# psql_splitter

a [`nom`](https://github.com/Geal/nom) parser combinator that matches a psql statement.

Postgres has a dialect of SQL that I'm going to call pgsql.

Postgres also has a client, `psql`, which recognizes meta-commands and inline-copy syntax that normal pgsql does not.

This parser recognizes psql meta-commands and pgsql tokens, allowing the `statement` combinator to know when a statement has ended.

The `statement` combinator also attaches color commentary on the end of a line to the previous statement, so `select 1; -- take a number` is matched as one statement.

## Status

This is a hack with some tests, a README, and a license.

This parser doesn't understand all the tokens in pgsql and psql, which has its pros:

- `psql_splitter::statement` is more likely to split statements out of invalid psql

- a simpler parser is easier to maintain

but also its cons:

- a simpler parser is more likely to be missing something about the enormous pgsql language

## Scope/features

- the splitter should match invalid psql, always splitting the input into one or more statements

- the splitter should treat psql meta-commands as statement terminators, e.g `select 1 as foo \gset` and `\d my_table` would both be complete statements.

- the splitter should ignore statement-terminators inside pgsql strings and parentheses and exotic psql such as `\if ... \endif` and `COPY table_name FROM STDIN; ... \.`

- the splitter should somewhat-reliably split psql and pgsql statements as defined by postgres, but not neccessarily other sql-like dialects derived from postgres's psql and pgsql.

- the splitter should be correct. Speed, a small binary, and memory efficiency aren't goals, but they're nice to have.

## Contributing

If you find a bug, please submit an issue with a reproduction case.

PRs for tooling would also be welcome.

## License

While this library is offered under the attached BSD 3-clause at ./LICENSE, `./src/*.sql` are excerpts from the postgres codebase, and so are covered by [the postgres license](https://www.postgresql.org/about/licence/).

## Prior art

[`libpg_query`'s `split_with_scanner`](https://github.com/pganalyze/libpg_query)

[`codeschool/sqlite-parser`'s test-suite-extraction tools](https://github.com/codeschool/sqlite-parser/blob/master/test/misc/test-grammar.pegjs)