| The current design is intended to separate Haddock from GHC as much as | possible, but putting documentation in .hi files would be going in the | opposite direction. There would have to be a compelling reason to do | that, something that we couldn't do another way. Actually it would in many ways be easier to put Haddock stuff in .hi files. But since GHC writes the .hi file, that would essentially mean merging GHC and Haddock into a single compile-and-documentation-generator. That would be cool in a way -- for example, GHCi would natively have access to the Haddock docs for a function. As Simon M says, the big reason not to do that is because it couples together two large projects, making each harder to develop independently. And that's a pretty big reason. Simon | -----Original Message----- | From: ghc-devs [mailto:ghc-devs-bounces@haskell.org] On Behalf Of Simon | Marlow | Sent: 19 March 2014 11:40 | To: Mateusz Kowalczyk; ghc-devs@haskell.org | Subject: Re: Haddock strings in .hi files | | On 18/03/2014 18:20, Mateusz Kowalczyk wrote: | > Hi all, | > | > I saw https://ghc.haskell.org/trac/ghc/ticket/5467 pop up in my inbox | > and it reminded me of something I've been wondering for a while: why | > do we not store Haddock docstrings in the interface file? | > | > I think that if we did, we could do some great things: | > | > 1. Show docs in GHCi (I vaguely recall someone working on this ~1 year | > ago, does anyone have any info?) | > | > 2. Allow Haddock to work a lot faster: the big majority of time spent | > when creating documentation is actually spent by Haddock calling | > various GHC functions, such as type-checking the modules. Only a small | > amount of time is actually spent by Haddock on other tasks such as | > parsing or outputting the documentation. If we could simply get | > everything we need from the .hi files, we save ourselves a lot of | time. | | Don't you still have to run the renamer at least? And in GHC, renaming | is tied up with typechecking, so it's hard to do one without the other. | Furthermore, if there is a missing type signature it's useful to be | able to put the inferred type in the documentation. I think I'm missing | the point somewhere - how does putting docs in the .hi file let you | avoid typechecking? | | I'm not really sure I see the benefit. If Haddock provided a library | that we can call from GHCi to get documentation, then we could show | documentation in GHCi. | | The current design is intended to separate Haddock from GHC as much as | possible, but putting documentation in .hi files would be going in the | opposite direction. There would have to be a compelling reason to do | that, something that we couldn't do another way. | | Cheers, | Simon | | | > 3. Allow Haddock to create partial documentation: a complaint I | > sometimes hear is if anything at all in the project doesn't type | > check, we don't get any documentation at all. I think that it'd be | > viable to generate only the documentation for the modules/functions | > that do type-check and perhaps skip type signatures for everything | else. | > | > Points 1. and 2. are of clear benefit. Point 3. is a simple | > afterthought and thinking about it some more, I think that maybe it'd | > be possible to do this with what we have right now: is type-checking | > separate parts of the module supported? Can we retrieve documentation | > for the parts that don't type-check? | > | > I am asking for input on what people think. I am not familiar at all | > with what goes into the .hi file (and I can't find anything concrete! | > Am I missing some wiki page?) at all and why. At the very least, 1. | > should be easy to implement. | > | > It was suggested that I submit a proposal for this as part of GSoC, | > namely implementing 1. and 2.. I admit that having much faster | > documentation builds would be amazing and Edward K. and Carter S. seem | > to think that this is very do-able in the 3 month period that GSoC | > runs over. | > | > While I say all this, I have already submitted my proposal on a | > different topic. I am considering writing this up and submitting this | > as well but I am looking for some insight into the problem first. | > | > If there are any students around still looking for ideas, please do | > speak up if you want to snatch this. If there are people that are | > eager to mentor something like this then I suppose they should speak | up too. | > | > Thanks! | > | _______________________________________________ | ghc-devs mailing list | ghc-devs@haskell.org | http://www.haskell.org/mailman/listinfo/ghc-devs

On 2014-03-20 at 11:39:47 +0100, Mateusz Kowalczyk wrote:

[...] the functionality to grab documentation for something can be provided by Haddock and GHCi then can use that. I think an extra benefit of this is that we get to decide in Haddock how we present such documentation: we can have different pretty printers if we so desire. If the strings are in .hi files themselves, the only choice is to show them verbatim which can be sub-par.

So if having documentation in GHCi is the sole goal, I think it's not worth putting the strings in the .hi file.

Just wondering: how would GHCi have to handle the case of e.g. ':reload' (in "-fobject-code" as well as "-fbyte-code" mode), in order to have access to up-to-date Haddock strings?

One strong reason for considering at least including the haddocks in the .hi files is build times. Currently if you have cabal configured to build and document every package running hackage requires you to recompile your entire source tree a second time to get information that we just dropped on the floor before spitting out the .hi file. For most of the users of GHC this is a 50% difference in compile times if they have cabal configured to generate haddocks. GHC doesn't have to understand the haddocks any more than it does now to support it, just include the content. Haddock could then just go through and load the .hi files rather than starting from scratch with parsing and typechecking the entire module, running template-haskell, just to get at the documentation. Any pythonesque :doc command support to me would be gravy. The reason I care at all is the build times. I regularly lose minutes out of each build just to regenerate docs and wind up skipping building them as much as I can get away with to avoid he pain. -Edward On Thu, Mar 20, 2014 at 4:08 AM, Simon Peyton Jones wrote:

| The current design is intended to separate Haddock from GHC as much as | possible, but putting documentation in .hi files would be going in the | opposite direction. There would have to be a compelling reason to do | that, something that we couldn't do another way.

Actually it would in many ways be easier to put Haddock stuff in .hi files. But since GHC writes the .hi file, that would essentially mean merging GHC and Haddock into a single compile-and-documentation-generator. That would be cool in a way -- for example, GHCi would natively have access to the Haddock docs for a function.

As Simon M says, the big reason not to do that is because it couples together two large projects, making each harder to develop independently. And that's a pretty big reason.

Simon

| -----Original Message----- | From: ghc-devs [mailto:ghc-devs-bounces@haskell.org] On Behalf Of Simon | Marlow | Sent: 19 March 2014 11:40 | To: Mateusz Kowalczyk; ghc-devs@haskell.org | Subject: Re: Haddock strings in .hi files | | On 18/03/2014 18:20, Mateusz Kowalczyk wrote: | > Hi all, | > | > I saw https://ghc.haskell.org/trac/ghc/ticket/5467 pop up in my inbox | > and it reminded me of something I've been wondering for a while: why | > do we not store Haddock docstrings in the interface file? | > | > I think that if we did, we could do some great things: | > | > 1. Show docs in GHCi (I vaguely recall someone working on this ~1 year | > ago, does anyone have any info?) | > | > 2. Allow Haddock to work a lot faster: the big majority of time spent | > when creating documentation is actually spent by Haddock calling | > various GHC functions, such as type-checking the modules. Only a small | > amount of time is actually spent by Haddock on other tasks such as | > parsing or outputting the documentation. If we could simply get | > everything we need from the .hi files, we save ourselves a lot of | time. | | Don't you still have to run the renamer at least? And in GHC, renaming | is tied up with typechecking, so it's hard to do one without the other. | Furthermore, if there is a missing type signature it's useful to be | able to put the inferred type in the documentation. I think I'm missing | the point somewhere - how does putting docs in the .hi file let you | avoid typechecking? | | I'm not really sure I see the benefit. If Haddock provided a library | that we can call from GHCi to get documentation, then we could show | documentation in GHCi. | | The current design is intended to separate Haddock from GHC as much as | possible, but putting documentation in .hi files would be going in the | opposite direction. There would have to be a compelling reason to do | that, something that we couldn't do another way. | | Cheers, | Simon | | | > 3. Allow Haddock to create partial documentation: a complaint I | > sometimes hear is if anything at all in the project doesn't type | > check, we don't get any documentation at all. I think that it'd be | > viable to generate only the documentation for the modules/functions | > that do type-check and perhaps skip type signatures for everything | else. | > | > Points 1. and 2. are of clear benefit. Point 3. is a simple | > afterthought and thinking about it some more, I think that maybe it'd | > be possible to do this with what we have right now: is type-checking | > separate parts of the module supported? Can we retrieve documentation | > for the parts that don't type-check? | > | > I am asking for input on what people think. I am not familiar at all | > with what goes into the .hi file (and I can't find anything concrete! | > Am I missing some wiki page?) at all and why. At the very least, 1. | > should be easy to implement. | > | > It was suggested that I submit a proposal for this as part of GSoC, | > namely implementing 1. and 2.. I admit that having much faster | > documentation builds would be amazing and Edward K. and Carter S. seem | > to think that this is very do-able in the 3 month period that GSoC | > runs over. | > | > While I say all this, I have already submitted my proposal on a | > different topic. I am considering writing this up and submitting this | > as well but I am looking for some insight into the problem first. | > | > If there are any students around still looking for ideas, please do | > speak up if you want to snatch this. If there are people that are | > eager to mentor something like this then I suppose they should speak | up too. | > | > Thanks! | > | _______________________________________________ | ghc-devs mailing list | ghc-devs@haskell.org | http://www.haskell.org/mailman/listinfo/ghc-devs _______________________________________________ ghc-devs mailing list ghc-devs@haskell.org http://www.haskell.org/mailman/listinfo/ghc-devs

My knowledge of precisely how haddock works is somewhat fuzzy in that it arises from a series of discussions a couple of years back. My observation was mostly that I run 'cabal install' it goes through all the modules building my .hi files, etc. Then I run cabal haddock and it spends all that time redoing the same work, just to go through and get at some information that we had right up until the moment we finished building. I'm not wedded to bolting the information into the .hi files being the solution, but the idea that we could avoid redoing that work is tantalizing. I'm mostly trying to avoid redoing all the same work twice in the build cycle of the average user. If there is an alternative strategy, such as, oh, I don't know, making haddock able to hook in plugin-style late as we're generating the .hi file to spit out what it needs to something else and interrogate/rename/whatever it needs the rest of the GHC API I'd be totally open that as well. -Edward On Thu, Mar 20, 2014 at 12:18 PM, Mateusz Kowalczyk

wrote:

...

One strong reason for considering at least including the haddocks in the .hi files is build times.

Currently if you have cabal configured to build and document every

On 20/03/14 16:08, Edward Kmett wrote: package

...

running hackage requires you to recompile your entire source tree a second time to get information that we just dropped on the floor before spitting out the .hi file.

For most of the users of GHC this is a 50% difference in compile times if they have cabal configured to generate haddocks.

GHC doesn't have to understand the haddocks any more than it does now to support it, just include the content.

Haddock could then just go through and load the .hi files rather than starting from scratch with parsing and typechecking the entire module, running template-haskell, just to get at the documentation.

Any pythonesque :doc command support to me would be gravy.

The reason I care at all is the build times. I regularly lose minutes out of each build just to regenerate docs and wind up skipping building them as much as I can get away with to avoid he pain.

-Edward

As Simon M points out, we still have to run the renamer which seems to be tightly bound with the type-checker. Where do you suggest the sizeable performance increase would be coming from in this case? For all the existing packages, we already read the docs from .haddock files so there's no difference there. For new packages we have to type-check and generate .haddock anyway so there's no difference there either.

It's not really about GHC having to know more about Haddock, it's about Haddock having to use GHC anyway, whether the strinsg are embedded or not.

-- Mateusz K.

Ok, I buy the argument that if we're already compiling everything, we shouldn't have to re-typecheck it all in Haddock. Of course if you're *not* already compiling everything, then the argument doesn't apply: Haddock does support generating documentation from source files without precompiling them, but I think if you ask the GHC API to load modules with -fno-code it should do the right thing: load up the .hi files if they're up to date, or typecheck the modules otherwise. So I think having GHC spit out the docs as a side-effect of compilation is fine, so long as we don't have to do all the Haddock processing inside GHC itself, and provided this eliminates Haddock's own interface files (which are a pain). If the docs go in the .hi file, then they must go in a separate section that is lazy parsed - we already do this for various other sections in the .hi file. I don't think this is easy, but it's probably doable. The code that attached docs to declarations is currently part of Haddock itself, so perhaps this has to move into GHC. Cheers, Simon On 20/03/2014 16:41, Edward Kmett wrote:

My knowledge of precisely how haddock works is somewhat fuzzy in that it arises from a series of discussions a couple of years back.

My observation was mostly that I run 'cabal install' it goes through all the modules building my .hi files, etc. Then I run cabal haddock and it spends all that time redoing the same work, just to go through and get at some information that we had right up until the moment we finished building.

I'm not wedded to bolting the information into the .hi files being the solution, but the idea that we could avoid redoing that work is tantalizing. I'm mostly trying to avoid redoing all the same work twice in the build cycle of the average user.

If there is an alternative strategy, such as, oh, I don't know, making haddock able to hook in plugin-style late as we're generating the .hi file to spit out what it needs to something else and interrogate/rename/whatever it needs the rest of the GHC API I'd be totally open that as well.

-Edward

On Thu, Mar 20, 2014 at 12:18 PM, Mateusz Kowalczyk mailto:fuuzetsu@fuuzetsu.co.uk> wrote:

On 20/03/14 16:08, Edward Kmett wrote: > One strong reason for considering at least including the haddocks in the > .hi files is build times. > > Currently if you have cabal configured to build and document every package > running hackage requires you to recompile your entire source tree a second > time to get information that we just dropped on the floor before spitting > out the .hi file. > > For most of the users of GHC this is a 50% difference in compile times if > they have cabal configured to generate haddocks. > > GHC doesn't have to understand the haddocks any more than it does now to > support it, just include the content. > > Haddock could then just go through and load the .hi files rather than > starting from scratch with parsing and typechecking the entire module, > running template-haskell, just to get at the documentation. > > Any pythonesque :doc command support to me would be gravy. > > The reason I care at all is the build times. I regularly lose minutes out > of each build just to regenerate docs and wind up skipping building them as > much as I can get away with to avoid he pain. > > -Edward > >

As Simon M points out, we still have to run the renamer which seems to be tightly bound with the type-checker. Where do you suggest the sizeable performance increase would be coming from in this case? For all the existing packages, we already read the docs from .haddock files so there's no difference there. For new packages we have to type-check and generate .haddock anyway so there's no difference there either.

It's not really about GHC having to know more about Haddock, it's about Haddock having to use GHC anyway, whether the strinsg are embedded or not.

-- Mateusz K.