'Trouble building conda recipe for python/c++ package

This is my first conda recipe, so please bear with me.

I'm trying to create a conda package for the tool NeuSomatic (https://github.com/bioinform/neusomatic). This package requires a build, but to run you just call python files on the command line, e.g. (from repo Readme):

python train.py \
    --candidates_tsv work_train/dataset/*/candidates*.tsv \
    --out work_train \
    --num_threads 10 \
    --batch_size 100

What I want to have the recipe do, is be able to just call the separate python packages from the command line after installation, e.g.

neusomatic-train \
    --candidates_tsv work_train/dataset/*/candidates*.tsv \
    --out work_train \
    --num_threads 10 \
    --batch_size 100

I thought I could get this right by using entry-points in the build (see meta.yaml example below), but these python files get some arguments in the if __name__ == "__main__" section (see https://github.com/bioinform/neusomatic/blob/master/neusomatic/python/train.py for example).

What am I doing wrong here?

meta.yaml:

{% set version = "0.2.1" %}

package:
  name: neusomatic
  version: {{ version }}

about:
  home: https://github.com/bioinform/neusomatic
  license: Creative Commons Attribution-NonCommercial-ShareAlike 4.0 International
  summary: NeuSomatic is based on deep convolutional neural networks for accurate somatic mutation detection. 
           With properly trained models, it can robustly perform across sequencing platforms, strategies, and 
           conditions. NeuSomatic summarizes and augments sequence alignments in a novel way and incorporates 
           multi-dimensional features to capture variant signals effectively. It is not only a universal but 
           also accurate somatic mutation detection method.

source:
    fn: v{{ version }}.tar.gz
    url: https://github.com/bioinform/neusomatic/archive/refs/tags/v{{ version }}.tar.gz
    md5: d315e16e825e1fbad71bfd7331ab9c3c

build:
  number: 1
  skip: False

requirements:
  build:
    - cmake =3.13.2

  entry_points:
    - neusomatic-preprocess = neusomatic.preprocess
    - neusomatic-train = neusomatic.train
    - neusomatic-postprocess = neusomatic.postprocess
    - neusomatic-call = neusomatic.call

  run:
    - python >=3.7
    - pytorch =1.1.0
    - torchvision =0.3.0
    - pybedtools =0.8.0
    - pysam =0.15.2
    - zlib =1.2.11
    - numpy =1.15.4
    - scipy =1.2.0
    - imageio =2.5.0
    - biopython =1.73
    - cudatoolkit =9.0
    - tabix =0.2.6
    - bedtools =2.27.1
    - samtools =1.9

build.sh:

#!/bin/bash
set -e
DIR="$( cd "$( dirname "${BASH_SOURCE[0]}" )" && pwd -P)"

rm -rf $DIR/third_party/SeqLib/ $DIR/third_party/seqan/ $DIR/neusomatic/build
pushd $DIR/neusomatic
  mkdir build
    pushd build
      cmake ..
      make
    popd
popd
pythonbashbuildcondabioinformatics


Solution 1:[1]

I managed to find a solution, by using some bash trickery in the build.sh script:

build.sh:

#!/bin/bash

set -euo pipefail
DIR="$( cd "$( dirname "${BASH_SOURCE[0]}" )" && pwd -P)"

rm -rf $DIR/third_party/SeqLib/ $DIR/third_party/seqan/ $DIR/neusomatic/build
pushd $DIR/neusomatic
  mkdir build
    pushd build
      cmake ..
      make
    popd
popd

### SOLUTION BELOW ###

outdir=$PREFIX/share/$PKG_NAME-$PKG_VERSION-$PKG_BUILDNUM
mkdir -p $outdir
mkdir -p $PREFIX/bin

cp -R neusomatic $outdir/neusomatic

ls -l $outdir

ln -s $outdir/neusomatic/python/preprocess.py $PREFIX/bin/neusomatic-preprocess
ln -s $outdir/neusomatic/python/train.py $PREFIX/bin/neusomatic-train
ln -s $outdir/neusomatic/python/call.py $PREFIX/bin/neusomatic-call
ln -s $outdir/neusomatic/python/postprocess.py $PREFIX/bin/neusomatic-postprocess

chmod 0755 "${PREFIX}/bin/neusomatic-preprocess"
chmod 0755 "${PREFIX}/bin/neusomatic-train"
chmod 0755 "${PREFIX}/bin/neusomatic-call"
chmod 0755 "${PREFIX}/bin/neusomatic-postprocess"

meta.yaml:

{% set version = "0.2.1" %}

package:
  name: neusomatic
  version: {{ version }}

about:
  home: https://github.com/bioinform/neusomatic
  license: Creative Commons Attribution-NonCommercial-ShareAlike 4.0 International
  summary: NeuSomatic is based on deep convolutional neural networks for accurate somatic mutation detection. 
           With properly trained models, it can robustly perform across sequencing platforms, strategies, and 
           conditions. NeuSomatic summarizes and augments sequence alignments in a novel way and incorporates 
           multi-dimensional features to capture variant signals effectively. It is not only a universal but 
           also accurate somatic mutation detection method.

source:
    fn: v{{ version }}.tar.gz
    url: https://github.com/bioinform/neusomatic/archive/refs/tags/v{{ version }}.tar.gz
    md5: d315e16e825e1fbad71bfd7331ab9c3c

build:
  number: 1
  skip: False

requirements:
  build:
    - cmake =3.13.2
  run:
    - python >=3.7
    - pytorch =1.1.0
    - torchvision =0.3.0
    - pybedtools =0.8.0
    - pysam =0.15.2
    - zlib =1.2.11
    - numpy =1.15.4
    - scipy =1.2.0
    - imageio =2.5.0
    - biopython =1.73
    - cudatoolkit =9.0
    - tabix =0.2.6
    - bedtools =2.27.1
    - samtools =1.9

test:
  commands:
    - neusomatic-preprocess --help 2> /dev/null
    - neusomatic-train --help 2> /dev/null
    - neusomatic-call --help 2> /dev/null
    - neusomatic-postprocess --help 2> /dev/null

Sources

This article follows the attribution requirements of Stack Overflow and is licensed under CC BY-SA 3.0.

Source: Stack Overflow

Solution Source
Solution 1 ljc

Related Questions