Why does GHC consider the LHS *syntactically* when inlining?

Question

According to the GHC docs:

...GHC will only inline the function if it is fully applied, where \"fully applied\" means applied to as many arguments a

Answer 1

This is a good question. I read through the Secrets of the Glasgow Haskell Compiler Inliner paper for some clues, but didn't find much.

Here's a hand-wavy explanation. GHC actually at times takes comp1 to comp2 -- which it calls "eta expansion". See this thread for some details: http://www.haskell.org/pipermail/glasgow-haskell-users/2011-October/020979.html

There also is (or was) a problem where this eta expansion can subtly change strictness. See this commit to the docs (which doesn't seem to be in the current ones, so either they haven't been rebuilt, or this has been fixed, not sure which): http://permalink.gmane.org/gmane.comp.lang.haskell.cvs.ghc/57721

In any case the above thread has SPJ explaining why we typically want to go in that direction whenever possible. So to deliberately go in the other direction in order to improve inlining seems a bit silly. As the secrets paper discusses, inlining promiscuously is not the greatest idea -- making the pragma even more of a blunt hammer so that functions got inlined whether or not it made sense to do so would probably hurt more than help overall, not to mention increase code bloat, since modules would have to keep the different levels of eta-shifted functions around all at once.

Anyway, as someone very much not a core GHC dev, that's what seems most likely to me.