shared-mime-info (PR#8278)

Hi,

On 03 Nov 2005 12:41:53 +0100, Peter Dalgaard <p.dalgaard at biostat.ku.dk> wr=
ote:

ripley at stats.ox.ac.uk writes:

We do not usually put features in R which are specific to just some
distributions of some OSes, and in this case to one editor on those.
We do not for example include the ESS mode for the much-more-widely-use=

d

Emacs family of editors.

This looks as if it might be appropriate to the Linux binary packages f=

or

R, so I suggest you contact their maintainers.  But my understanding is
that this is an issue for gedit and not for R.  Indeed .R is just a
convention (one of many choices, including .r and .q) for R itself.

I do wonder why you concentrated on .R files and not .Rd files, where I
find syntax highlighting more useful.

Mime-types shouldn't be distribution-specific or even editor-specific,
should they? The whole point is that they can be used for things like
email attachments that pass from one OS to the other.

It might be useful to have the mime-type definitions for R (and Rd)
files centralized in R core, with the appropriate OS conventions
systematized. But I think we need to know more. Who keeps track of
mime-types? Can we just grab text/x-R (and text/x-Rd and
application/x-Rdata)? To which extent the XML format a standard; is it
only used by particular applications?

As far as I know, at least in Debian, the mimetypes are tracked by
shared-mime-info package. The upstream is freedesktop.org. I do not
know about oficial standarts, but Gnome and KDE tries to adher to some
of the freedesktop.org standarts. I can confirm that mimetypes
provided by shared-mime-info are widely used in Gnome, for some time
now.

One further thought about this:

On SUSE,

rpm -qif /usr/share/mime/

points at

http://www.freedesktop.org/wiki/Software_2fshared_2dmime_2dinfo

So I guess that the proper tree to bark at is the upstreams
maintainers of

http://freedesktop.org/~jrb/shared-mime-info-*.tar.gz

Instructions there say to submit new XML files to

https://bugs.freedesktop.org/buglist.cgi?product=shared-mime-info&bug_status=NEW&bug_status=ASSIGNED&bug_status=REOPENED

It would likely be a good idea to send them first to R-devel for review.

Hi,

On 04 Nov 2005 13:51:56 +0100, Peter Dalgaard <p.dalgaard at biostat.ku.dk> wrote:

One further thought about this:

On SUSE,

rpm -qif /usr/share/mime/

points at

http://www.freedesktop.org/wiki/Software_2fshared_2dmime_2dinfo

So I guess that the proper tree to bark at is the upstreams
maintainers of

http://freedesktop.org/~jrb/shared-mime-info-*.tar.gz

Instructions there say to submit new XML files to

https://bugs.freedesktop.org/buglist.cgi?product=shared-mime-info&bug_status=NEW&bug_status=ASSIGNED&bug_status=REOPENED

It would likely be a good idea to send them first to R-devel for review.

I already barked at upstream. The upstream barked back. The result is here:

https://bugs.freedesktop.org/show_bug.cgi?id=1782

There you can find xml file for R scripts. I've made it from some
example. It is really only a proof of a concept. But it would not be
very difficult to produce xml files for mimetypes of all R related
files. We must only decide which R related files would benefit from
having mimetypes.

My proposal is
1. R source code, R scripts. Files with extensions .R, .r and others
(.q?, .s?, .S?). Mimetypes text/x-R, text/x-Rsrc

2. R documentation files. File extension .Rd. Mimetype text/x-Rd

3. RData files. File extension .RData, files which at beginning have
RDX2. Mimetype application/x-RData.

4. Rhistory files. File extension .Rhistory. Mimetype text/x-Rhistory

5. R transcript files from ESS/Emacs. File extension .Rt. Mimetype
text/x-Rtranscript

The relevant xml code could be pushed upstream to end up in
freedesktop.org.xml, or it could be distributed with R linux package,
and installed into relevant subdirectory of /usr/share/mime. With a
bit more work the result could be, that people using for example
Nautilus (graphical Gnome browser) could see R related files displayed
with R logo, and clicking them could result in various appropriate
actions. For example for .RData R process could be iinvoked and
relevant .RData file could be loaded.

I could write and test the xml code. But first we have to agree on
which files benefit from having mimetypes and how the mimetypes should
be named. Feel free to suggest.

Vaidotas Zemlys writes:

On 04 Nov 2005 13:51:56 +0100, Peter Dalgaard wrote:

[SNIP RPM discussion]

There you can find xml file for R scripts. I've made it from some
example. It is really only a proof of a concept. But it would not be
very difficult to produce xml files for mimetypes of all R related
files. We must only decide which R related files would benefit from
having mimetypes.

My proposal is
1. R source code, R scripts. Files with extensions .R, .r and others
(.q?, .s?, .S?). Mimetypes text/x-R, text/x-Rsrc

My inclination would be to stick with .R, possibly adding .r to guard
against Windows case-folding issues, but .r used to be Ratfor files.
.q/.s/.S are used by some people supporting both R and S-PLUS, but I
don't think they care how such files are displayed by Nautilus and
Konqueror...

2. R documentation files. File extension .Rd. Mimetype text/x-Rd

OK, modulo case-fold

3. RData files. File extension .RData, files which at beginning have
RDX2. Mimetype application/x-RData.

Why the RDX2 bit?? We do have .RDA from windows, too.

4. Rhistory files. File extension .Rhistory. Mimetype text/x-Rhistory

OK.

5. R transcript files from ESS/Emacs. File extension .Rt. Mimetype
text/x-Rtranscript

.Rout, please. Also .Rout.save and .Rout.fail. (And it's not just
ESS that creates them).

Also

6. Rprofile files .Rprofile or Rprofile.

I could write and test the xml code. But first we have to agree on
which files benefit from having mimetypes and how the mimetypes should
be named. Feel free to suggest.