Incorrect formatting of bun placeholders (e.g. ?1)

[ysmilda]

Describe the bug
In our sql files we make use of placeholders (?0) so that these can be replaced dynamically when read by bun. When ran through sqlfmt it adds a space between the ? and the 0. The following statement "time" = ?0 becomes "time" = ? 0 which makes it invalid.

This is not a feature of standard SQL but I would expect the formatter to leave these statements as is. My assumption is that the ? is read as a statement of it's own, but I can't locate the source of it.

To Reproduce
See example above

Expected behavior
I would expect no formatting to happen between these characters.

Additional context
sqlfmt, version 0.14.3

[tconbeer]

Similar to psycopg placeholders #198 and pg variables/placeholders a5d7dd0

Looks like bun's syntax is r"\?\d*", including the bare ?, which collides with the use of ? as an operator in some dialects (we're parsing ? as an operator now). e.g., postgres: jsonb ? text → boolean

But I think we can easily support r"\?\d+" in the other_identifiers rule without any bad side-effects. Change would be here:

sqlfmt/src/sqlfmt/rules/core.py

Lines 134 to 144 in 3a2b998

	Rule(
	name="other_identifiers",
	priority=600,
	pattern=group(
	r"@\w+", # stages
	r"\$\d+", # pg placeholders
	r"\$\w+", # variables
	r"%(\([^%()]+\))?s", # psycopg placeholders
	),
	action=partial(actions.add_node_to_buffer, token_type=TokenType.NAME),
	),

[ysmilda]

That indeed did the trick, see #358