Try ShortText or Basement.String instead of Text for strings in Expr

[sjakobi]

See http://hackage.haskell.org/package/text-short-0.1.2/docs/Data-Text-Short.html#t:ShortText
and http://hackage.haskell.org/package/basement-0.0.10/docs/Basement-String.html#t:String.

Both are UTF8 which could be a benefit.

[sjakobi]

For reference, there was a slightly related discussion on string interning on discourse

[ocharles]

Is there anyway we could get that archive into our Discourse instance?

[Gabriella439]

@ocharles: Not that I know of

[ocharles]

Presumably you are repyling to "Is there any automated way we could that archive into our Discourse instance?" which isn't quite what I'm asking. We could certainly manually restore the backup by going through this archive, assuming Discourse gives us a way to manually create posts or something. Maybe not worth doing now though...

[sjakobi]

Is there anyway we could get that archive into our Discourse instance?

You can find some threads via this Google search: https://www.google.de/search?q=site%3Adiscourse.dhall-lang.org/t/

Then click on the green arrow next to each URL to access the cache.

[ocharles]

Sorry to derail this issue, but maybe we should mention that on Discourse itself. Possibly pinned, though maybe that's not important.

[sjakobi]

Sorry to derail this issue, but maybe we should mention that on Discourse itself.

See https://discourse.dhall-lang.org/t/sorry-i-deleted-the-forum-history/11/6

Let's continue this discussion to Discourse.

[sjakobi]

I've tried converting the core data structures in Dhall.Core and Dhall.Eval to use ShortText instead of Text on the sjakobi/1032-text-short branch.

So far I've only seen tiny speedups for normalizing cpkg's pkg-set.dhall. The deep-nested-large-record benchmark got slightly slower. It's possible that the benefits would be more pronounced on very large projects.

Since many of the libraries that dhall talks to use Text, there's a bit of additional conversion overhead. Which is sad, as many of these libraries eventually convert their UTF16 Text to UTF8 again… Let's hope that text-utf8 eventually takes off!