darcs patch: Add liftField and maybeField to ParseUti... (and 1 more)

Bertram Felgenhauer

16 Jul 2006 16 Jul '06

2:06 p.m.

The following two patches add the possibility to install hidden packages. The first one adds functionality to convert between StanzaFields of different types and uses that to simplify the parsing and pretty printing code for package descriptions. The patch should not change the behaviour in any way. The second patch is quite simple and just adds a flag. It would be much more involved without the first patch. What do you think? regards, Bertram Sun Jul 16 19:26:06 CEST 2006 Bertram Felgenhauer * Add liftField and maybeField to ParseUtils and use them to simplify PackageDescription. This gets rid of the special handling of 'exposed-modules' and simplifies some pretty printing and parsing code. It'll make adding new library fields much easier. Sun Jul 16 19:35:08 CEST 2006 Bertram Felgenhauer * Add 'exposed' library field to allow installing hidden packages.

Attachments:

add-liftfield-and-maybefield-to-parseutils-and-use-them-to-simplify-packagedescription_.dpatch (text/x-darcs-patch — 19.0 KB)

Show replies by date

Bertram Felgenhauer

27 Jul 27 Jul

8:05 p.m.

Hi, I'd really appreciate some feedback on this, did anybody look at it? I believe the first patch could help Cabal even if installing packages hidden will not become a (Distribution.Simple) Cabal feature. regards, Bertram Original message: http://www.haskell.org//pipermail/cabal-devel/2006-July/000035.html

Bertram Felgenhauer

10:56 p.m.

Hi, I've discussed the patch with Duncan Coutts. The result is the attached revised patch. Changes: - fix the handling of 'x-' fields - with the previous patch cabal would complain about these as unknown in most contexts - document liftField and maybeField and make their use for converting stanzas more obvious in PackageDescription.hs The second patch is unchanged. A note on installing packages hidden: This should be useful for programs that provide their own plugins or other libraries for internal use. It seems to be simple enough for inclusion in Distribution.Simple. regards, Bertram

Duncan Coutts

28 Jul 28 Jul

5:57 a.m.

New subject: darcs patch: Add liftField and maybeField to ParseUti... (and 1 more)

On Fri, 2006-07-28 at 04:56 +0200, Bertram Felgenhauer wrote:

...

Hi,

I've discussed the patch with Duncan Coutts. The result is the attached revised patch. Changes:

- fix the handling of 'x-' fields - with the previous patch cabal would complain about these as unknown in most contexts - document liftField and maybeField and make their use for converting stanzas more obvious in PackageDescription.hs

Thanks. Isaac: I'm happy with both of these patchs if you are. I've also got a slight improvement on the docs for liftField which I'll add if these go in.

...

The second patch is unchanged. A note on installing packages hidden: This should be useful for programs that provide their own plugins or other libraries for internal use. It seems to be simple enough for inclusion in Distribution.Simple.

Note that it'd also be useful to be able to Cabalise the GHC API package and the old hslibs packages (which makes bundling them with Haskell implementations easier as people who care can easily get at them still). Duncan

Simon Marlow

1 Aug 1 Aug

7:23 a.m.

New subject: darcs patch: Add liftField and maybeField to ParseUti... (and 1 more)

Duncan Coutts wrote:

...

On Fri, 2006-07-28 at 04:56 +0200, Bertram Felgenhauer wrote:

...
Hi,

I've discussed the patch with Duncan Coutts. The result is the attached revised patch. Changes:

- fix the handling of 'x-' fields - with the previous patch cabal would complain about these as unknown in most contexts - document liftField and maybeField and make their use for converting stanzas more obvious in PackageDescription.hs

Thanks.

Isaac: I'm happy with both of these patchs if you are.

I've also got a slight improvement on the docs for liftField which I'll add if these go in.

Dammit. I've done a lot of refactoring of the parsing stuff, including doing exactly what Bertram did with liftField, but I think I've pushed the changes further. If this patch goes in I have some serious merging to do. I suppose this is my own fault for taking so long with this configuration stuff... Cheers, Simon

Duncan Coutts

2 Aug 2 Aug

2:11 p.m.

New subject: darcs patch: Add liftField and maybeField to ParseUti... (and 1 more)

On Tue, 2006-08-01 at 12:23 +0100, Simon Marlow wrote:

...

Duncan Coutts wrote:

...

...
Isaac: I'm happy with both of these patchs if you are.

I've also got a slight improvement on the docs for liftField which I'll add if these go in.

Dammit. I've done a lot of refactoring of the parsing stuff, including doing exactly what Bertram did with liftField, but I think I've pushed the changes further. If this patch goes in I have some serious merging to do.

We can hold off, that's no problem. I'm sure it'll be easier to merge in any remaining of Bertram's changes after you push your config stuff than the other way around. Duncan

Isaac Jones

1 Aug 1 Aug

2:26 p.m.

Duncan Coutts writes:

...

On Fri, 2006-07-28 at 04:56 +0200, Bertram Felgenhauer wrote:

...
Hi,

I've discussed the patch with Duncan Coutts. The result is the attached revised patch. Changes:

- fix the handling of 'x-' fields - with the previous patch cabal would complain about these as unknown in most contexts

Right, it shouldn't ever complain / warn about x-fields, right?

...

...
- document liftField and maybeField and make their use for converting stanzas more obvious in PackageDescription.hs

Sounds good.

...

Thanks.

Isaac: I'm happy with both of these patchs if you are.

I've also got a slight improvement on the docs for liftField which I'll add if these go in.

Cool. I'm OK with it, but worried about Simon's patch that does the same thing. Simon, can you tease out that patch in short order and compare it side-by-side w/ Bertram's? Should we just go ahead and apply Bertram's?

...

...
The second patch is unchanged. A note on installing packages hidden: This should be useful for programs that provide their own plugins or other libraries for internal use. It seems to be simple enough for inclusion in Distribution.Simple.

Note that it'd also be useful to be able to Cabalise the GHC API package and the old hslibs packages (which makes bundling them with Haskell implementations easier as people who care can easily get at them still).

So this is done w/ a field rather than a flag to configure, but it sounds more like a configure option in most cases, no? Shouldn't this be up to the user more than the package author? It should be a pretty rare package that's meant to be hidden by default. We could always have both a field and a configure flag. peace, isaac

Bertram Felgenhauer

8:01 p.m.

Isaac Jones wrote:

...

Duncan Coutts writes:

...
...
- fix the handling of 'x-' fields - with the previous patch cabal would complain about these as unknown in most contexts

Right, it shouldn't ever complain / warn about x-fields, right?

Indeed. I had missed that parseBInfoField handled those.

...

Cool. I'm OK with it, but worried about Simon's patch that does the same thing. Simon, can you tease out that patch in short order and compare it side-by-side w/ Bertram's? Should we just go ahead and apply Bertram's?

I'm happy either way. It's not an urgent change, it just seemed to be the right thing to do, in order to add new library stanzas.

...

So this is done w/ a field rather than a flag to configure, but it sounds more like a configure option in most cases, no? Shouldn't this be up to the user more than the package author? It should be a pretty rare package that's meant to be hidden by default.

I wanted this for lambdabot - it's a stand-alone application, but it's plugin-based; one of the plugin's utilities (runplugs) has an auxillary module (ShowQ) that's arguably not useful for general use. To install it properly, a hidden package seemed to be the right thing. And after refactoring the parsing of the package description adding support for that to Distribution.Simple is a very simple change to Cabal.

...

We could always have both a field and a configure flag.

That makes sense to me, actually. regards, Bertram

Duncan Coutts

2 Aug 2 Aug

2:22 p.m.

New subject: darcs patch: Add liftField and maybeField to ParseUti... (and 1 more)

On Tue, 2006-08-01 at 11:26 -0700, Isaac Jones wrote:

...

Duncan Coutts writes:

...

...
...
- fix the handling of 'x-' fields - with the previous patch cabal would complain about these as unknown in most contexts

Right, it shouldn't ever complain / warn about x-fields, right?

Right.

...

...
Isaac: I'm happy with both of these patchs if you are.

I've also got a slight improvement on the docs for liftField which I'll add if these go in.

Cool. I'm OK with it, but worried about Simon's patch that does the same thing. Simon, can you tease out that patch in short order and compare it side-by-side w/ Bertram's? Should we just go ahead and apply Bertram's?

I'm happy to wait and apply after Simon.

...

...
Note that it'd also be useful to be able to Cabalise the GHC API package and the old hslibs packages (which makes bundling them with Haskell implementations easier as people who care can easily get at them still).

So this is done w/ a field rather than a flag to configure, but it sounds more like a configure option in most cases, no? Shouldn't this be up to the user more than the package author?

I don't think so. It is actually the package author who knows if it *must* be installed hidden, eg if it severely pollutes the module namespace like hslibs. It's true that the user might want to make it hidden even if the package author didn't, though at the moment we don't provide the user with the option and no one complains. They can change it using ghc-pkg anyway.

...

It should be a pretty rare package that's meant to be hidden by default.

That's true, though there are hslibs, the GHC api and lambdabot/Yi/hIDE style plug-in packages. Duncan

6866

Age (days ago)

6883

Last active (days ago)

List overview

Download

8 comments

4 participants

participants (4)

Bertram Felgenhauer
Duncan Coutts
Isaac Jones
Simon Marlow

darcs patch: Add liftField and maybeField to ParseUti... (and 1 more)

tags

participants (4)