feature request: possibility of outputting just a few trees

[odanoburu]

while writing a new lin, it is nice to able to get fast feedback on the results. gftest can be somewhat slow depending on the tested function (:/), so having a way of getting preliminary results fast would be nice -- specially if easy to implement! if these trees look fine, one can then proceed to doing the full test.

[inariksit]

Is this for just the normal mode with -f SlowFunction? For -o, I actually have this hack in place https://github.com/GrammaticalFramework/gftest/blob/master/src/Main.hs#L396-L397 , for getting the tested categories to flash by. Checking /tmp/ also works to see that it is creating bunch of trees and testing them.

The function testFun returns a full string with all results; this is in order to easier decide whether to write to stdout or to a file. If you want, you could change testFun into an IO function; instead of this unlines, make it into mapM_ print.

Another thing, if you're not doing it already, definitely compile your PGF with the GF flag --optimize-pgf! Basque doesn't even run on my computer without that.

[odanoburu]

Is this for just the normal mode with -f SlowFunction? For -o, I actually have this hack in place

yes, but I could use -o too! are you using strictness to make it run faster?

Another thing, if you're not doing it already, definitely compile your PGF with the GF flag --optimize-pgf!

I'm doing it! it should probably be done by default when --make is on, given how dramatic a difference it makes!

[inariksit]

are you using strictness to make it run faster?

I have no clue how to use strictness stuff in Haskell :-P that thing about printing to a file was just a random poking around and "huh this seems to work, let's keep it".

it should probably be done by default when --make is on, given how dramatic a difference it makes!

Hmm, I was trying to find out if there are reasons why it is not default--here's something in Krasimir's PhD thesis pages 76-77: "The final conclusion is that the optimized and the unoptimized grammars are semantically different." I don't know if that's good enough reason to not have optimized version as default, but at least I learned something new today!