diff mbox

gnu: Add mash.

Message ID 87eg4urqq3.fsf@ike.i-did-not-set--mail-host-address--so-tickle-me
State New
Headers show

Commit Message

Marius Bakke Sept. 8, 2016, 10:06 p.m. UTC
Leo Famulari <leo@famulari.name> writes:

> On Thu, Sep 01, 2016 at 11:00:39AM +0100, Marius Bakke wrote:
>> I had these in inputs initially and was surprised to see no references.
>> Both seems to be compiled into the final program[0]: when running "mash
>> info" on an invalid file (the provided data/refseq.msh), a generic
>> capnproto exception is thrown (src/capnp/serialize.c++:159).
>
> I wonder, does using native-inputs work when building mash for another
> architecture?

That's interesting, the package indeed fails to build on i686.

Sketch.cpp:(.text+0xdf): undefined reference to `memcpy@GLIBC_2.2.5'

I don't understand why, the symbol versions should be the same, no?

Are there any clever linker flags we can throw at it, or is setting
supported-systems acceptable?

>> That raises another question: should the htslib and capnproto licenses
>> be listed too, since they are part of the binary output?
>
> Good question, I'm not sure. I'd guess "yes", along with a code comment
> explaining what's going on.

I've attached a patch below, with license comments and amd64 only.

>> I'm not a bioinformatician (just a mere sysadmin for such), but have
>> been going through the tutorial and things appear to work fine.
>
> Ah, bioinformatics software... all bets are off ;)

You haven't seen anything yet! This is the nice part of my queue ;)

Thanks!
Marius

Comments

Leo Famulari Sept. 10, 2016, 9:20 p.m. UTC | #1
On Thu, Sep 08, 2016 at 11:06:44PM +0100, Marius Bakke wrote:
> Leo Famulari <leo@famulari.name> writes:
> 
> > On Thu, Sep 01, 2016 at 11:00:39AM +0100, Marius Bakke wrote:
> >> I had these in inputs initially and was surprised to see no references.
> >> Both seems to be compiled into the final program[0]: when running "mash
> >> info" on an invalid file (the provided data/refseq.msh), a generic
> >> capnproto exception is thrown (src/capnp/serialize.c++:159).
> >
> > I wonder, does using native-inputs work when building mash for another
> > architecture?
> 
> That's interesting, the package indeed fails to build on i686.

Do you mean that it fails when you try to build from x86_64 using `guix
build --system=i686-linux`? I wondered if that particular case would
work since the packages that are native-inputs would only be built for
the architecture of the builder, if I understand correctly.
Leo Famulari Sept. 10, 2016, 9:42 p.m. UTC | #2
On Thu, Sep 08, 2016 at 11:06:44PM +0100, Marius Bakke wrote:
> * gnu/packages/bioinformatics.scm (mash): New variable.

Anyways, pushed as 84be3b9920120e7cc03095baca06d61b7f3fb741. If the
package needs more changes, we will change it :)
diff mbox

Patch

From 9e8102ed2d5bf9334e5311f2ac917aed2f451361 Mon Sep 17 00:00:00 2001
From: Marius Bakke <mbakke@fastmail.com>
Date: Tue, 30 Aug 2016 18:49:21 +0100
Subject: [PATCH] gnu: Add mash.

* gnu/packages/bioinformatics.scm (mash): New variable.
---
 gnu/packages/bioinformatics.scm | 57 +++++++++++++++++++++++++++++++++++++++++
 1 file changed, 57 insertions(+)

diff --git a/gnu/packages/bioinformatics.scm b/gnu/packages/bioinformatics.scm
index f34acd1..decca6c 100644
--- a/gnu/packages/bioinformatics.scm
+++ b/gnu/packages/bioinformatics.scm
@@ -76,6 +76,7 @@ 
   #:use-module (gnu packages python)
   #:use-module (gnu packages readline)
   #:use-module (gnu packages ruby)
+  #:use-module (gnu packages serialization)
   #:use-module (gnu packages statistics)
   #:use-module (gnu packages tbb)
   #:use-module (gnu packages tex)
@@ -3046,6 +3047,62 @@  sequences).")
               "http://mafft.cbrc.jp/alignment/software/license.txt"
               "BSD-3 with different formatting"))))
 
+(define-public mash
+  (package
+    (name "mash")
+    (version "1.1.1")
+    (source (origin
+              (method url-fetch)
+              (uri (string-append
+                    "https://github.com/marbl/mash/archive/v"
+                    version ".tar.gz"))
+              (file-name (string-append name "-" version ".tar.gz"))
+              (sha256
+               (base32
+                "08znbvqq5xknfhmpp3wcj574zvi4p7i8zifi67c9qw9a6ikp42fj"))
+              (modules '((guix build utils)))
+              (snippet
+               ;; Delete bundled kseq.
+               ;; TODO: Also delete bundled murmurhash and open bloom filter.
+               '(delete-file "src/mash/kseq.h"))))
+    (build-system gnu-build-system)
+    (arguments
+     `(#:tests? #f ; No tests.
+       #:configure-flags
+       (list
+        (string-append "--with-capnp=" (assoc-ref %build-inputs "capnproto"))
+        (string-append "--with-gsl=" (assoc-ref %build-inputs "gsl")))
+       #:make-flags (list "CC=gcc")
+       #:phases
+       (modify-phases %standard-phases
+         (add-after 'unpack 'fix-includes
+           (lambda _
+             (substitute* '("src/mash/Sketch.cpp" "src/mash/CommandFind.cpp")
+               (("^#include \"kseq\\.h\"")
+                "#include \"htslib/kseq.h\""))
+             #t))
+         (add-before 'configure 'autoconf
+           (lambda _ (zero? (system* "autoconf")))))))
+    (native-inputs
+     `(("autoconf" ,autoconf)
+       ;; Capnproto and htslib are statically embedded in the final
+       ;; application. Therefore we also list their licenses, below.
+       ("capnproto" ,capnproto)
+       ("htslib" ,htslib)))
+    (inputs
+     `(("gsl" ,gsl)
+       ("zlib" ,zlib)))
+    (supported-systems '("x86_64-linux"))
+    (home-page "https://mash.readthedocs.io")
+    (synopsis "Fast genome and metagenome distance estimation using MinHash")
+    (description "Mash is a fast sequence distance estimator that uses the
+MinHash algorithm and is designed to work with genomes and metagenomes in the
+form of assemblies or reads.")
+    (license (list license:bsd-3          ; Mash
+                   license:expat          ; HTSlib and capnproto
+                   license:public-domain  ; MurmurHash 3
+                   license:cpl1.0))))     ; Open Bloom Filter
+
 (define-public metabat
   (package
     (name "metabat")
-- 
2.9.3