The pgquery-scala library provides a set of bindings to access the power of libpg_query at compile time in your scala projects. Since Postgres' parser is used no need for a local container nor a database connection to check your code ! All errors won't be cought this way (for instance references to unexisting columns and such) but its overload on compilation time is minimal.
Currently the versions 15, 16 and 17 of Postgres are supported. Just add a line in the scripts/PG_VERSIONS file to add one !
The main engine here is pgquery.macros.check, which is loading the right library for the asked Postgres Version and is issuing the native calls.
To poke around and test the query parsing, you can use the provided convienience method pgquery.pgsql(PgVersion, String). For now, sql errors results in warnings but switching between warning and hard errors is in the roadmap.
pgquery.pgsql("select foo from bar where bloop = 2") // compiles fine
pgquery.pgsql("select foo form bar where bloop = 2") // errors at compile time pgquery-scala-doobie extends doobie-postgres with additional interpolators.
val age: Int = ???
pgsql"select name from user"
// Output a `Fragment` like the `sql` interpolator
pgsql"select name from user where age = $age"
// Support for interpolated values is limited : for now they are replaced with NULL for parsing
inline given PgVersion = PgVersion.V15
// default version is the latest (17) but you can tune it
pgsql15"merge into target using source on [...] returning [...]" // fails
pgsql17"merge into target using source on [...] returning [...]" // compiles
//Exemples: `returning` is not available on `merge` until version 17pgquery-scala-skunk extends skunk-core with additional interpolators
The universal pgsql interpolator is not available due du some interraction between pgquery-scala and skunk macros,
but the fixed SQL versions of them are available for testing.
NOTE: Maven publishing TBD. In the meantime, you have to clone and run sbt publishLocal.
NOTE: The set of libraries provided contains only macros, so their scope can be only Compile on SBT.
You can use the script provided in scripts/codegen.sh to generate the shared libraries for you :
PGQUERY_BUILD_BINDINGS=false PGQUERY_BUILD_SHARED=true
Since this is compiling many files from many version of libpg_query you can expect it to take time.
Then add the lib folder that got created to your LD_LIBRARY_PATH variable when running tooling, IDE, and such.
In order to help debugging nasty env variable issues the pgquery.utils.comptimEnv("SOME_ENV_VAR") is provided, just call it in the current file and the compiler should emit an INFO log.