Is it possible that your size pre-calculation is inadvertently inducing quadratic behavior? I remember a pretty-printer implementation that generated both a line-broken and single-line version as it went, in order to avoid quadratic behavior due to backtracking. (IIRC, it would discard one as it went, if it determined that the single-line version of a sub-structure wouldn’t fit.) I don’t remember the exact details though so I know that’s a bit vague. I know you are caching size in order to avoid calculating it multiple times, but I’m wondering if that is still inducing double-traversal (once for size, once to generate the final text), and if that double could turn into a square. Just a possibility. Jeff

On Apr 17, 2023, at 3:51 AM, Johannes Waldmann wrote:

Dear Cafe,

I was looking for a way to pretty-print Haskell literals (with lists, tuples, records with named and positional notation) like this example

( Leftist { tree = Branch { left = Branch { left = Leaf, key = 4, right = Leaf } , key = 3 , right = Leaf } , refs = listToFM [ ( Ref 13, [ 0 ] ), ( Ref 17, [ ] ) ] } , [ Any, Any ] )

for each sub-structure, the indentation level (for the following lines) should increase - by a _fixed_ amount. in the above example: line break after "tree = Branch". But (missing from this example), line break _before_ the list starts in "{ foo = [ 42 , ... ] ... }".

I found this impossible to do with wl-pprint but perhaps I did not try hard enough.

Instead, I "invented" combinators `nest` and `skip` and made this prototypical implementation https://gitlab.imn.htwk-leipzig.de/autotool/all0/-/blob/master/todoc/src/Tex... (it has some explanatory text at the top) see also https://gitlab.imn.htwk-leipzig.de/autotool/all0/-/issues/960

but certainly this cannot be a new idea.

While I do like the semantics (in the context of my application), I don't like the performance of my implementation. What am I doing wrong? It's just updating indentation level and current position, this should not take any time at all?

Of course, it would be best if I don't need the implementation at all - if the effect could be achieved via some combinators in established libraries (that have optimized implementation).

Any pointers appreciated.

Best regards - J. _______________________________________________ Haskell-Cafe mailing list To (un)subscribe, modify options or view archives go to: http://mail.haskell.org/cgi-bin/mailman/listinfo/haskell-cafe Only members subscribed via the mailman list are allowed to post.

Sorry I'm late to the party. I wanted something like that a long time ago, and I thought "this should be easy" so I wrote a "simple" library for it: https://github.com/elaforge/karya/blob/work/Util/Format.hs https://github.com/elaforge/karya/blob/work/Util/Pretty.hs The reason for the scare quotes is that it turned out to be much harder and more finicky than I thought. This implementation still has at least one formatting bug which I did track down at some point, but I believe it was a missing logical distinction which would be hard to fix. But in fact I've been using it for about 9 years and it's worked well enough. I haven't done any performance analysis, but I've never felt the need for it, and I've printed some pretty large things. So, I don't know if it's hard because it's actually hard, or because I didn't do a great design, or because I wasn't motivated to put further work into it after it got to good enough. On Mon, Apr 17, 2023 at 3:51 AM Johannes Waldmann wrote:

Dear Cafe,

I was looking for a way to pretty-print Haskell literals (with lists, tuples, records with named and positional notation) like this example

( Leftist { tree = Branch { left = Branch { left = Leaf, key = 4, right = Leaf } , key = 3 , right = Leaf } , refs = listToFM [ ( Ref 13, [ 0 ] ), ( Ref 17, [ ] ) ] } , [ Any, Any ] )

for each sub-structure, the indentation level (for the following lines) should increase - by a _fixed_ amount. in the above example: line break after "tree = Branch". But (missing from this example), line break _before_ the list starts in "{ foo = [ 42 , ... ] ... }".

I found this impossible to do with wl-pprint but perhaps I did not try hard enough.

Instead, I "invented" combinators `nest` and `skip` and made this prototypical implementation https://gitlab.imn.htwk-leipzig.de/autotool/all0/-/blob/master/todoc/src/Tex... (it has some explanatory text at the top) see also https://gitlab.imn.htwk-leipzig.de/autotool/all0/-/issues/960

but certainly this cannot be a new idea.

While I do like the semantics (in the context of my application), I don't like the performance of my implementation. What am I doing wrong? It's just updating indentation level and current position, this should not take any time at all?

Of course, it would be best if I don't need the implementation at all - if the effect could be achieved via some combinators in established libraries (that have optimized implementation).

Any pointers appreciated.

Best regards - J. _______________________________________________ Haskell-Cafe mailing list To (un)subscribe, modify options or view archives go to: http://mail.haskell.org/cgi-bin/mailman/listinfo/haskell-cafe Only members subscribed via the mailman list are allowed to post.