Re: pure parsers and reentrant scanners

Peter Eisentraut <peter@eisentraut.org>

From: Peter Eisentraut <peter@eisentraut.org>

To: Andreas Karlsson <andreas@proxel.se>, pgsql-hackers <pgsql-hackers@postgresql.org>

Date: 2024-12-19T12:48:24Z

Lists: pgsql-hackers

Commits

Same data as JSON: GET /api/v1/messages/:b64id/commits the thread's linked commits as JSON, with link sources. API reference →

Return yyparse() result not via global variable
- 473a575e0597 18.0 landed
Remove flex version checks
- 0869ea43e9c2 18.0 landed
Drop warning-free support for Flex 2.5.35
- 6fdd5d956343 18.0 landed
plpgsql: pure parser and reentrant scanner
- 7b27f5fd36cb 18.0 landed
flex code modernization: Replace YY_EXTRA_TYPE define with flex option
- b1ef48980ddd 18.0 landed
guc: reentrant scanner
- d663f150b5ed 18.0 landed
jsonpath scanner: reentrant scanner
- 2a7425d7eef9 18.0 landed
syncrep parser: pure parser and reentrant scanner
- db6856c9913f 18.0 landed
replication parser: pure parser and reentrant scanner
- e4a8fb8fefb9 18.0 landed
bootstrap: pure parser and reentrant scanner
- 3e4bacb17100 18.0 landed
Small whitespace improvement
- 399d0f1e11b5 18.0 landed
Prevent redeclaration of typedef yyscan_t
- 382092a0cd2c 18.0 landed
seg: pure parser and reentrant scanner
- 1f0de66ea2a5 18.0 landed
cube: pure parser and reentrant scanner
- 802fe923e3cd 18.0 landed

On 17.12.24 01:46, Andreas Karlsson wrote:
> On 12/16/24 8:39 AM, Peter Eisentraut wrote:
>> I'll leave it at this for now and wait for some reviews.
> 
> I really like this work since it makes the code cleaner to read on top 
> of paving the way for threading.
> 
> Reviewed the patches and found a couple of issues.
> 
> - Shouldn't yyext in syncrep_scanner_init() be allocated on the heap? Or 
> at least on the stack but by the caller?

I think it's correct the way it is.  It's only a temporary space for the 
scanner, so we can allocate it in the innermost scope.

> - I think you have flipped the parameters of replication_yyerror(), see 
> attached fixup patch.

Good catch.  There was also a similar issue with syncrep_yyerror().

> - Some white space issues fixed in an attached fixup patch.

committed

> - Also fixed the static remaining variables in the replication parser in 
> an attached patch.

Thanks, I'll take a look at that.

> - There seems to be a lot left to do to make the plpgsql scanner 
> actually re-entrant so I do not think it would makes sense to commit the 
> patch which sets the re-entrant option before that is done.

Ok, we can hold that one back until the full stack including the parser 
is done.