Update Dokchitser calculator to compiled version

[mraum]

Cohen provided a C-version of Dokchitser's script. This wraps that script and cleans up the script. There were several issues present and doctests for the elliptic curves have been adapted to correctly set the maximal imaginary value.

Apply:

1. trac-11024-dokchitser.patch​
2. trac-11024-dokchitser-2.patch​
3. trac-11024-dokchitser-bordeaux.patch​
4. ​http://sage.math.washington.edu/home/mraum/dokchitser-20110325.spkg

[cremona]

The patch and spkg installed OK (on 63-bit ubuntu) and I am now testing.

• The spkg needs a version number, so use the date (e.g. dokchitser-20110325.spkg)
• There needs to be an SPKG.txt file
• The spkg needs to be under mercurial version control (but *not* the src/ subdirectory).
• The spkg-install looks good!

These are trivialities, but necessary!

[mraum]

Updated spkg available.

[cremona]

I'll look at the new spkg. Meanwhile testing revealed one small problem:

sage -t "devel/sage-main/doc/en/bordeaux_2008/elliptic_curves.rst"

[cremona]

New spkg: SPKG.txt seems to refer to a different spkg!!

Also I suggest adding .hgignore to list src/

[mraum]

A tiny change to the Dokchitser wrapper. The old way of finding the library was kind of dump. Also the spkg now contains the .hgignore as you suggested.

[mraum]

Something went wrong with diff, this patch is better.

[mraum]

Fix the doctest in the Bordeaux notes. We need to set the maximal imaginary value correctly.

[jdemeyer]

Martin: patches should be created using hg export tip instead of hg diff. If you use queues, be sure to set a meaningful commit message using hg qrefresh -e. The commit message should contain the ticket number on the first line.

[rishi]

From my small tests, the patch works good, and it eliminated the bug which I saw yesterday. I would second Jeroen's request that there should be a meaningful commit message. Right now it is set of diffs.

I would have done it right now, but it will be hard for me to write an appropriate commit message without essentially reviewing it. I will do it next weekend if no one has done it by then.

[jdemeyer]

If you are adding a new spkg to Sage, you should also update various files in SAGE_ROOT, in particular spkg/install and spkg/standard/deps.

[mraum]

Jeroen added a comment to #11005, stating that it is probably better to add the single c-file to the Sage library. We probably also want to do this for the Dokchitser calculator.

I hope to have time for this soon.

[cremona]

Thanks a lot for reviving this!

If you fancy doing the same with the Simon 2-descent scripts that would be fantastic. Both Martin Raum and I tried to do that a couple of years ago but failed (and I don't remember why). Again, Henri Cohen and I first spent a while cleaning up the scripts to avoid global variables and such things. I would certainly be willing to work jointly on that as I know his code pretty well. There are also plans within the pari developers group to implement all that in the pari library, but I don't know how far that has got.

[chapoton]

Hello,

Just to be clear, I have not yet done anything serious for this ticket.

It seems that the current spkg (as given by the link) needs to be updated, because the file computel5.gp.c contains three calls to the function gtoser with a wrong number of arguments.

[cremona]

Karim Belabas says "This is scheduled for implementation in PARI core." I don't know how far off that is, but once it has happened we will not need the gp script but instead additions to the pari library interface.

[chapoton]

see ​https://pari.math.u-bordeaux.fr/dochtml/html/_L_minusfunctions.html

[chapoton]

see #26098 for another tentative

[pbruin]

It looks like this is no longer needed after #26098. Can someone who has worked on this ticket confirm this?

[cremona]

I agree with that (see my comment 20 above).

[pbruin]

Replying to cremona:

I agree with that (see my comment 20 above).

OK, thanks.

[chapoton]

Either one decides to remove the Dokschitser scripts and rely only on pari itself, or this ticket still makes sense. So far, both implementations coexist.

[cremona]

Replying to chapoton:

Either one decides to remove the Dokschitser scripts and rely only on pari itself, or this ticket still makes sense. So far, both implementations coexist.

Please let us remove the Dokchitser script. It is very old and certainly not supported by its author who mvoed over entirely to Magma for these computations.

I suppose some deprecation should be put in place first?