upgrade rpy2 to 2.7.5

[EmmanuelCharpentier]

The current version (2.3.8) of rpy2 packaged in Sage is too old to work with the current version of the Jupyter notebook

Manually upgrading to 2.7.4 (via sage -pip install --upgrade rpy2 seems to work. However in also upgrades six (to 1.10.0-py2.py3-non-any.whl (???) and singledispatch (but seems to use the current (cached) version).

I also get the following warning :

DEPRECATION: Uninstalling a distutils installed project (rpy2) has been deprecated and will be removed in a future version. This is due to the fact that uninstalling a distutils project will only partially uninstall the project.

which I do not really understand....

New tarballs attached ; you may also download rpy2, six and singledispatch directly from python.org.

Edit : version bumped to 2.7.5

CC: @kiwifb

Component: packages: standard

Keywords: r-project jupyter

Author: Emmanuel Charpentier

Branch: 50462d2

Reviewer: François Bissey, Travis Scrimshaw

Issue created by migration from https://trac.sagemath.org/ticket/19638

[EmmanuelCharpentier]

comment:1

Extract from this question on sage-devel :

My pip upgrade showed me that this upgrade entails :

• an upgrade to six, which is already packaged in Sagemath (no problem)
• installing the singledispatch package, which is NOT packaged.

ISTR that, on sage-devel, it has been decided that a new package should remain optional for at least a year before becoming a standard package. If we follow this policy, the standard package rpy2 would depend on the optional package singledispatch. This seems a bit incoherent...

Another solution would be to keep the standard rpy2 package at 2.3.8, and have a rpy2-newer optional, depending on singledispatch. This seems a bit contrived, to say the least. Furthermore, I do not know how to proceed...

A third solution would be to silently include singledispatch in the source to rpy2. This "solution" seems to guarantee future incompatibilities.

What should I do ?

Bonus question : how to package rpy2 to signal that it depends on singledispatch ?

[EmmanuelCharpentier]

Branch: u/charpent/upgrade_rpy2_to_2_7_4

[EmmanuelCharpentier]

Commit: e503295

[EmmanuelCharpentier]

Author: Emmanuel Charpentier

[EmmanuelCharpentier]

comment:3

This version does the job (%R and %%R run OK in the Jupyter notebook).

Not yet ready for review, because :

• No documentation (yet) of this new interface
• not yet ptestlong'ed (and won't be until late tonight)...

Any suggestion on how to document the Jupyter interface ?

[EmmanuelCharpentier]

Attachment: singledispatch-3.4.0.3.tar.gz

Tarball for the new sigledispatch package

[EmmanuelCharpentier]

Attachment: six-1.10.0.tar.gz

New tarball for six

[EmmanuelCharpentier]

New tarball for rpy2

[EmmanuelCharpentier]

comment:4

Attachment: rpy2-2.7.4.tar.gz

Our damn "corporate" firewall mangled the upload of new tarballs...

[EmmanuelCharpentier]

comment:5

Passes ptestlong with no error.

I still haven't the foggiest idea about where to document the Jupyter interface. We need to mention four things at the very minimum :

• Load it with %load_ext rpy2.ipython.
• Line magic : %R (may take args to pass data from/to Sage).
• Cell magic : {{{%%R}} (idem).
• Help is available online.

By the way, the Jupyter notebook doesn't seem documented in the Sage documentation.

[kiwifb]

comment:6

We don't put tarball in attachment anymore, we rarely put any attachment anymore actually. We usually put a pointer to where to download the source in the description, see for example #10295.

[EmmanuelCharpentier]

comment:7

Links to the original tarballs added in the description.

I'm still waiting for comments on the right place to add documentation.

[kiwifb]

comment:8

I honestly don't know where to document it in this case. You may have to try to stir things up on sage-devel.

The warning you get from pip is nothing to worry about. If I am not mistaken it is just complaining that you are uninstalling/upgrading something that wasn't installed with pip and that support for this will be discontinued (and rightly so, there is no sane way for pip to know it is doing a good job with the removal. pip as a package manager actually tracks the file installed, not so for stuff that has been installed with plain distutils).

[sagetrac-git]

Changed commit from e503295 to b4e9773

[sagetrac-git]

Branch pushed to git repo; I updated commit sha1. New commits:

b4e9773

Embryonic documentation for the RPy2 interface.

[EmmanuelCharpentier]

comment:10

Passes ptestlong when added to a pristine 6.10rc0 git clone.

Embryonic documentation added as an addendum to the r interface documentation.

It appears that, while useful in its present state, this work should be extended to a (re-)implementation of the R interface (our present implementation uses expect).

I shall question sage-devel and open another ticket for this.

Still requiring review.

[kiwifb]

Reviewer: François Bissey

[kiwifb]

comment:12

Good enough for me. I have been using rpy-2.7.x since October in sage-on-gentoo without a hitch and 2.7.4 in particular since the end of November.

[kiwifb]

comment:13

2.7.5 is out since we are now aiming for sage-6.11 I am ok to re-review it after a revision bump.

[EmmanuelCharpentier]

New tarball for rpy2-2.7.5

[sagetrac-git]

Changed commit from b4e9773 to 50462d2

[sagetrac-git]

comment:14

Attachment: rpy2-2.7.5.tar.gz

Branch pushed to git repo; I updated commit sha1 and set ticket back to needs_review. New commits:

50462d2

Upgrade rpy2 to rpy2-2.7.5.

[EmmanuelCharpentier]

comment:15

Passes ptestlong.

[tscrim]

comment:16

I have verified this.

[EmmanuelCharpentier]

comment:17

Replying to @tscrim:

I have verified this.

Thanks. You should add your (real) name to the "Reviewers" field...

[kiwifb]

Changed reviewer from François Bissey to François Bissey, Travis Scrimshaw

[vbraun]

comment:19

Please don't attach tarballs to trac tickets, they just end up in our trac database forever

[vbraun]

Changed branch from u/charpent/upgrade_rpy2_to_2_7_4 to 50462d2

[kcrisman]

Changed commit from 50462d2 to none

[kcrisman]

Changed reviewer from François Bissey, Travis Scrimshaw to François Bissey, Travis Scrimshaw