Making a package for Primitive Permutation Groups Library

[olexandr-konovalov]

This library (http://www.gap-system.org/Datalib/prim.html) should become a proper package, loadable via LoadPackage. The library is in the separate prim directory, and the documentation is already in GAPDoc format at http://www.gap-system.org/Manuals/doc/ref/chap50.html#X7AE00EA7791F2574.

For the beginning, I suggest to make it needed package for GAP for backwards compatibility (so GAP will not start without it). Packages that rely on it (CTblLib and IRREDSOL, maybe more) would then be able to include it in their dependencies. Eventually we will put it into the list of packages loaded by default but not needed to start GAP.

Questions:

• License: GPL v2?
• Authors: http://www.gap-system.org/Datalib/prim.html says "Colva Roney-Dougal and many others", and gives Colva's and @hulpke details for contacts. Who should be the authors of the package?
• package name: "prim" or "PrimitiveGroups" or "PrimitiveGroupsLibrary"? My current choice is "PrimitiveGroups" as the name of the directory (note that LoadPackage("prim");) will still work as long as there is no other package started with "prim").

I am happy to do the conversion which is pretty straightforward. I am setting GAP 4.9.0 as a milestone and assigning it to myself.

[olexandr-konovalov]

This should now incorporate #818 which extends it for degrees up to 4095

[dimpase]

What kind of backwards compatibility are you talking about? GAP 4.8.* starts and runs without this library just fine:

% gap
┌───────┐ GAP 4.8.3, 19-Mar-2016, build of 2016-05-16 05:03:46 (UTC)
│ GAP │ http://www.gap-system.org
└───────┘ Architecture: x86_64-unknown-linux-gnu-gcc-default64
Libs used: gmp, readline
Loading the library and packages ...
Packages: GAPDoc 1.5.1
Try '?help' for help. See also '?copyright' and '?authors'
gap>

Please do not make things needed, if they are not...

[olexandr-konovalov]

@dimpase: missing trans causes errors while loading Browse package, and missing prim - errors for CTblLib and IRREDSOL (not surprising, since package authors of course assume that these are parts of the core system). They will need time to adapt their packages and add new "TransitiveGroups" and "PrimitiveGroups" packages to their dependencies.

[dimpase]

Why don't you instead just add these dependencies to these packages?
Or send them (trivial) patches to apply...

Instead you somehow prefer to intentionally break GAP.

[markuspf]

On Wed, Jun 15, 2016 at 05:01:48AM -0700, Dima Pasechnik wrote:

Why don't you instead just add these dependencies to these packages?
Or send them (trivial) patches to apply...

Instead you somehow prefer to intentionally break GAP.

I think the way Alex suggests how to go about making these components packages
is absolutely sensible. We make a change, and don't break backwards
compatibility for a while.

In addition we can send those patches, or suggest to people that they should
update their packages.

[dimpase]

@markuspf : Alex suggests to break backwards compatibility for people who do not need trans and prim, by co-opting the latter into "GAP core".

[olexandr-konovalov]

@dimpase I don't understand how this could be "co-opting". They became independent packages, so taken out of the GAP core where they are now, and nothing will change for people who do not need trans and prim at all.

[dimpase]

And, by the way, it looks more like a bug in Browse that it needs trans to load, for I cannot imagine Browse using trans functionality for its inner workings; similarly CTblLib probably does not really need prim.

[dimpase]

@alex-konovalov : for your own version of "core" you might be right. But trans and prim are data libraries, no more than that.

[fingolfin]

While I disagree with @dimpase rhetorics, I also don't see why we should make the new packages "needed" -- wouldn't it be sufficient to have them autoload by default? I.e. always load them if they are present, but if they are not, just don't?

Eventually, they would then be moved to the regular default list of autoloaded packages, which user's can overwrite. We'd do that after all relevant packages had time to adapt (i.e. to add explicit dependencies to the new trans / prim packages).

It's important to do these things incremental, to allow a smooth transition.

[fingolfin]

However, let me point out that right now, we do consider trans and prim to be parts of GAP core. Yes, you can remove them and GAP still works. And if you wantede to, you could also remove support for, say, permutation groups from GAP, and GAP could still work (it would be a lot more effort to achieve that, of course). So what? Doing it is not supported. Moving them into packages is what will make it "supported", so I'd think you should be happy about this...

That said, even if we marked those new packages as "needed", then people who want to hack GAP by removing these GAP core parts can simply hack GAP a little more, and edit the bits where we mark the packages as needed.

But again, I don't think we need to mark them "needed".

[olexandr-konovalov]

@dimpase : neither of this is a bug.

• Browse provides BrowseTransitiveGroupsInfo which shows the contents of the transitive groups library: http://www.gap-system.org/Manuals/pkg/Browse/doc/chapA.html#X832846F486391385
• ctbllib/gap4/ctdbattr.g uses PRIMRANGE, AllTransitiveGroups and AllPrimitiveGroups to setup mappings between the character table library and these group libraries
• IRREDSOL package relies on DegreeOfMatrixGroup defined in prim/irredsol.gd.