I wrote:
> Noah Misch wrote:
>>> During ANALYZE, in analyze.c, functions compute_minimal_stats
>>> and compute_scalar_stats, values whose length exceed
>>> WIDTH_THRESHOLD (= 1024) are not used for calculating statistics
>>> other than that they are counted as "too wide rows" and assumed
>>> to be all different.
>>>
>>> This works fine with regular tables;
>>> With foreign tables the situation is different. Even though
>>> values exceeding WIDTH_THRESHOLD won't get used, the complete
>>> rows will be fetched from the foreign table. This can easily
>>> exhaust maintenance_work_mem.
>>> I can think of two remedies:
>>> 1) Expose WIDTH_THRESHOLD in commands/vacuum.h and add documentation
>>> so that the authors of foreign data wrappers are aware of the
>>> problem and can avoid it on their side.
>>> This would be quite simple.
>> Seems reasonable. How would the FDW return an indication that a
value was
>> non-NULL but removed due to excess width?
>
> The FDW would return a value of length WIDTH_THRESHOLD+1 that is
> long enough to be recognized as too long, but not long enough to
> cause a problem.
Here is a simple patch for that.
Yours,
Laurenz Albe