> > > Having said
> > > that it seems that another solution would be to check for duplicated constants
> > > in fill_in_constant_length
> >
> > Yes, I started thinking along these lines as well, where we check for
> > duplicates
> > in fill_in_constant_length; or after jumbling, we de-duplicate locations if we
> > have a squashable list, which is what I have attached with updated test cases.
> >
> > This means we do need to scan the locations one extra time during jumbling,
> > but I don't see that as a problem. Maybe there is another better way?
>
> Why? What I hand in mind is something like this, after a quick test it seems to
> be able to address both the original case and the one you've discovered.
ahh, what you have works because clocations is sorted by location before
the loop starts in `fill_in_constant_lengths` , so comparing locations
makes sense in this case. I am OK with this.
```
/*
* Sort the records by location so that we can process them in order while
* scanning the query text.
*/
if (jstate->clocations_count > 1)
qsort(jstate->clocations, jstate->clocations_count,
sizeof(LocationLen), comp_location);
```
I was really hoping that the fix could be inside of query jumbling, as I think
pg_stat_statements or any consumers of query jumbling don't need to
care more about squashing ( or other internals of jumbling ).
So now I also wonder if we should also move:
```
static char *generate_normalized_query(JumbleState *jstate, const char *query,
int query_loc, int *query_len_p);
static void fill_in_constant_lengths(JumbleState *jstate, const char *query,
int query_loc);
```
from pg_stat_statements.c to queryjumblefuncs.c?
--
Sami Imseih
Amazon Web Services (AWS)
