Re: Character expansion with ICU collations

On 11.06.21 22:05, Finnerty, Jim wrote:
>>>
> You can have these queries return both rows if you use an
> accent-ignoring collation, like this example in the documentation:
>
> CREATE COLLATION ignore_accents (provider = icu, locale =
> 'und-u-ks-level1-kc-true', deterministic = false);
> <<
>
> Indeed. Is the dependency between the character expansion capability and accent-insensitive collations documented anywhere?

The above is merely a consequence of what the default collation elements
for 'ß' are.

Expansion isn't really a relevant concept in collation. Any character
can map to 1..N collation elements. The collation algorithm doesn't
care how many it is.

> Can a CI collation be ordered upper case first, or is this a limitation of ICU?

I don't know the authoritative answer to that, but to me it doesn't make
sense, since the effect of a case-insensitive collation is to throw away
the third-level weights, so there is nothing left for "upper case first"
to operate on.

> More generally, is there any interest in leveraging the full power of ICU tailoring rules to get whatever order someone may need, subject to the limitations of ICU itself? what would be required to extend CREATE COLLATION to accept an optional sequence of tailoring rules that we would store in the pg_collation catalog and apply along with the modifiers in the locale string?

yes