Missing parameters in the arg_walker callback of raw_expression_tree_walker ?

[daamien]

Hi !

This may not be a bug but it is at least very weird :)

I'm trying to use the raw_expression_tree_walker function, but I don't understant why the arg_walker callback function as no parameters... It's signature is

unsafe extern C fn() -> bool

https://docs.rs/pgrx/latest/pgrx/pg_sys/fn.raw_expression_tree_walker.html

This callback seems useless if I can't inspect the node itself...

I would expect the walker callback signature to be

unsafe extern C fn(* mut Node, * c_void) -> bool

Looking at the example below in the PostgreSQL code, I see that the tree_walker_callback used in raw_expression_tree_walker does have 2 parameters....

https://github.com/postgres/postgres/blob/master/src/backend/parser/analyze.c#L3576

Am I missing something obvious ?

[workingjubilee]

@daamien Mostly that Postgres 16 diffs this: postgres/postgres@1c27d16

[workingjubilee]

@daamien I am happy to accept a PR that does the relevant mischief to make all versions conform to the interface that Postgres 16 describes via the appropriate function-pointer casting/transmutes.

[daamien]

I'm afraid this is way beyond my capabilities :/

[daamien]

Hi @workingjubilee

As I'm moving forward with this, I realize that PR #1596 does cover query_tree_walker, expression_tree_walker and raw_expression_tree_walker but there's a few more functions that are also concerned by PG16's signature change

• query_or_expression_tree_walker
• range_table_entry_walker
• planstate_tree_walker
• range_table_walker

As far as I can tell, with their current signature those 4 functions a pretty much useless at the moment.

Now I guess I should be able to copycat what you did on #1596 and submit a patch to fix those 4 functions.... Should I make a single PR for them all or a distinct PR for each of them ?

[workingjubilee]

@daamien Feel free to take all of them in one.