Ricardo Wurmus <ricardo.wur...@mdc-berlin.de> skribis:
> From 8829683fffc03dec7f2faecea75cdd7831ce1741 Mon Sep 17 00:00:00 2001
> From: Ricardo Wurmus <ricardo.wur...@mdc-berlin.de>
> Date: Wed, 16 Dec 2015 14:45:28 +0100
> Subject: [PATCH] import: Add Bioconductor importer and updater.
>
> * guix/import/cran.scm (bioconductor->guix-package,
> %bioconductor-updater, latest-bioconductor-release,
> bioconductor-package?): New procedures.
> (%bioconductor-url, %bioconductor-svn-url): New variables.
> (description->package): Update signature to distinguish between packages
> from different repositories.
> (latest-release): Rename procedure ...
> (latest-cran-release): ... to this.
> (cran-package?): Do not assume all R packages are available on CRAN.
> * tests/cran.scm: Update tests.
> * guix/scripts/import/bioconductor.scm: New file.
> * guix/scripts/import.scm (importers): Add "bioconductor" importers.
> * guix/scripts/refresh.scm (%updaters): Add "%bioconductor-updater".
> * doc/guix.texi: Document Bioconductor importer and updater.
[...]
> @item cran
> the updater for @uref{http://cran.r-project.org/, CRAN} packages;
> +@item bioconductor
> +the updater for @uref{http://www.bioconductor.org/, Bioconductor} packages;
“R packages”
> - (properties ,`(,'quasiquote ((,'upstream-name . ,name))))
> + (properties ,`(,'quasiquote ((,'upstream-name . ,name)
> + (,'r-repository . ,repository))))
What about adding ‘upstream-name’ only when the upstream name is
different from the Guix name minus “r-”?
Regarding ‘r-repository’, I guess it can be inferred from the source
URL?
The risk with properties is that they are missing from current packages,
they are easily forgotten (since there’s no expansion-time check for
these), and forgetting them would lead to packages being silently
ignored by ‘guix refresh’.
> (define (cran-package? package)
> "Return true if PACKAGE is an R package from CRAN."
> - ;; Assume all R packages are available on CRAN.
> - (string-prefix? "r-" (package-name package)))
> + ;; Assume all R packages are available on CRAN, unless otherwise indicated
> + ;; by the r-repository property.
> + (let ((properties (package-properties package)))
> + (and (string-prefix? "r-" (package-name package))
> + (or (not properties)
> + (not (assoc-ref properties 'r-repository))
> + (eqv? 'cran (assoc-ref properties 'r-repository))))))
We could check whether the source URL starts with “mirror://cran”, no?
And forget about the ‘r-repository’ property?
> +(define (bioconductor-package? package)
> + "Return true if PACKAGE is an R package from Bioconductor."
> + (let ((properties (package-properties package)))
> + (and (string-prefix? "r-" (package-name package))
> + properties
> + (eqv? 'bioconductor (assoc-ref properties 'r-repository)))))
Likewise, should we check based on the URL?
> +++ b/guix/scripts/import/bioconductor.scm
I was wondering whether this should be a separate script or not. A
precedent would be the ELPA importer, which takes a repository name as
an argument. We could have done the same with the CRAN importer here.
WDYT?
The rest LGTM.
Thank you!
Ludo’.
