On Sun, Mar 17, 2019 at 6:03 PM Tom Lane <tgl@sss.pgh.pa.us> wrote:
> Andrew Dunstan <andrew.dunstan@2ndquadrant.com> writes:
> > Why are we installing the jsonpath_gram.h file? It's not going to be
> > used by anything else, is it? TBH, I'm not sure I see why we're
> > installing the scanner.h file either.
>
> As near as I can see, jsonpath_gram.h and jsonpath_scanner.h exist only
> for communication between jsonpath_gram.y and jsonpath_scan.l. Maybe
> we'd be better off handling that need by compiling the .l file as part
> of the .y file, as we used to do with the core lexer and still do with
> several others (cf comments for commit 72b1e3a21); then we wouldn't
> even have to generate these files much less install them.
>
> A quick look at jsonpath_scan.c shows that it's pretty innocent of
> the tricks we've learned over the years for flex/bison portability;
> in particular I see that it's #including <stdio.h> before postgres.h,
> which is a no-no. So that whole area needs more review anyway.
Attached patch is getting rid of jsonpath_gram.h. Would like to see
results of http://commitfest.cputube.org/ before committing, because
I'm not available to test this of windows machine. There would be
further patches rearranging jsonpath_gram.y and jsonpath_scan.l.
------
Alexander Korotkov
Postgres Professional: http://www.postgrespro.com
The Russian Postgres Company