Browse files

Adding new command: planemo dependency_script

This is a squashed commit of pull request #310 for issue #303,
for converting tool_dependencies.xml install recipes into bash

There is a lot that could be done better, or added, including:

- refactor to use a visitor pattern instead of my .to_bash() methods
- expand the command line API with options for paths and filenames
- setting defaults like download cache via ~/.planemo.yml
- complete the action coverage (especially the R/Python/Perl environments)
- avoid collisions in the download cache which currently assumes unique filenames

However, this is enough to help with automating dependency
installation in a continuous integration setup like TravisCI.
  • Loading branch information...
peterjc committed Oct 7, 2015
1 parent c857676 commit f798c7e29b2276ce68b828e72fc6a6460c73792b
Showing with 596 additions and 5 deletions.
  1. +235 −0 planemo/commands/
  2. +12 −0 planemo/
  3. +265 −5 planemo/shed2tap/
  4. +84 −0 tests/
@@ -0,0 +1,235 @@
import os
import sys

import click

from xml.etree import ElementTree as ET

from import info, error
from planemo.cli import pass_context
from planemo import options

from planemo.shed2tap.base import BasePackage, Dependency

# We're using strict bash mode for
preamble_dep_install = """#!/bin/bash
#set -euo pipefail
set -eo pipefail
if [[ ! -d $INSTALL_DIR ]]
echo "ERROR: Environment variable INSTALL_DIR not a directory!"
exit 1
# Set full strict mode now, side stepping case $INSTALL_DIR not setup.
set -euo pipefail
if [[ ! -d $DOWNLOAD_CACHE ]]
echo "Using $DOWNLOAD_CACHE for cached downloads."
export INSTALL_DIR=`(cd "$INSTALL_DIR"; pwd)`
echo "Using $INSTALL_DIR for the installed files."
# Create a randomly named temp folder for working in
(umask 077 && mkdir $dep_install_tmp) || exit 1

final_dep_install = """echo "Cleaning up..."
rm -rf $dep_install_tmp
echo "======================"
echo "Installation complete."
echo "======================"

# Expect user to "source" so don't set strict mode,
# and don't use the exit command!
preamble_env_sh = """#!/bin/bash
if [[ ! -d $INSTALL_DIR ]]
echo "ERROR: Environment variable INSTALL_DIR not a directory!"
export INSTALL_DIR=`(cd "$INSTALL_DIR"; pwd)`

def find_tool_dependencis_xml(path, recursive):
"""Iterator function, quick & dirty tree walking."""
if os.path.isfile(path):
if os.path.basename(path) == "tool_dependencies.xml":
yield path
elif os.path.isdir(path):
p = os.path.join(path, "tool_dependencies.xml")
if os.path.isfile(p):
yield p
if recursive:
for f in sorted(os.listdir(path)):
p = os.path.join(path, f)
if os.path.isdir(p):
# TODO: yield from
for x in find_tool_dependencis_xml(p, recursive):
yield x

def convert_tool_dep(dependencies_file):
"""Parse a tool_dependencies.xml into and commands.
Returns two lists of strings, commands to add to and respectively.
install_cmds = []
env_cmds = []

root = ET.parse(dependencies_file).getroot()
package_els = root.findall("package")

packages = []
dependencies = []
for package_el in package_els:
install_els = package_el.findall("install")
assert len(install_els) in (0, 1)
if len(install_els) == 0:
repository_el = package_el.find("repository")
assert repository_el is not None, "no repository in %s" % repository_el
dependencies.append(Dependency(None, package_el, repository_el))
install_el = install_els[0]
packages.append(BasePackage(None, package_el, install_el, readme=None))

if not packages:
info("No packages in %s" % dependencies_file)
return [], []

assert len(packages) == 1, packages
package = packages[0]
name = package_el.attrib["name"]
version = package_el.attrib["version"]

# TODO - Set $INSTALL_DIR in the script
# os.environ["INSTALL_DIR"] = os.path.abspath(os.curdir)
for action in package.all_actions:
inst, env = action.to_bash()

if install_cmds:
install_cmds.insert(0, 'cd $dep_install_tmp')
install_cmds.insert(0, 'specifc_action_done=0')
install_cmds.insert(0, 'echo "%s"' % ('=' * 60))
install_cmds.insert(0, 'echo "Installing %s version %s"' % (name, version))
install_cmds.insert(0, 'echo "%s"' % ('=' * 60))
if env_cmds:
env_cmds.insert(0, 'specifc_action_done=0')
env_cmds.insert(0, '#' + '=' * 60)
env_cmds.insert(0, 'echo "Setting environment variables for %s version %s"' % (name, version))
env_cmds.insert(0, '#' + '=' * 60)
# TODO - define $INSTALL_DIR here?

return install_cmds, env_cmds

def process_tool_dependencies_xml(tool_dep, install_handle, env_sh_handle):
"""Writes to handles, returns success as a boolean."""
if not os.path.isfile(tool_dep):
error('Missing file %s' % tool_dep)
return False
if not os.stat(tool_dep).st_size:
error('Empty file %s' % tool_dep)
return False
install, env = convert_tool_dep(tool_dep)
except Exception as err:
# TODO - pass in ctx for logging?
error('Error processing %s - %s' %
(click.format_filename(tool_dep), err))
if not isinstance(err, (NotImplementedError, RuntimeError)):
# This is an unexpected error, traceback is useful
import traceback
error(traceback.format_exc() + "\n")
return False
# Worked...
for cmd in install:
install_handle.write(cmd + "\n")
for cmd in env:
env_sh_handle.write(cmd + "\n")
return True

def cli(ctx, paths, recursive=False, fail_fast=True, download_cache=None):
"""Prepare a bash shell script to install tool requirements (**Experimental**)
An experimental approach parsing tool_dependencies.xml files into
bash shell scripts, intended initially for use within Continuous
Integration testing setups like TravisCI.
Parses the specified ``tool_dependencies.xml`` files, and converts them into
an installation bash script (default ````), and a shell script
(default ````) defining any new/edited environment variables.
These are intended to be used via ``bash`` (once), and as
``source`` prior to running any of the dependencies to set the
environment variable within the current shell session.
Both ```` and ```` require ``$INSTALL_DIR`` be defined
before running them, set to an existing directory with write permissions.
Beware than if run on multiple tools, they can over-write each other (for
example if you have packages for different versions of the same tool). In
this case make separate calls to ``planemo dependency_script`` and call
the scripts with different installation directories.
This command will download (and cache) any URLs specified via Galaxy
download actions. This is in order to decompress them and determine the
relevant sub-folder to change into as per the Tool Shed install mechanism,
so that this can be recorded as a ``cd`` comand in the bash script.
The download cache used by ``planemo dependency_script`` and the resulting
output script ```` defaults to ``./download_cache`` (under
the current working directory), and can be set with ``$DOWNLOAD_CACHE``.
This is experimental, and is initially intended for use within continuous
integration testing setups like TravisCI to both verify the dependency
installation receipe works, and to use this to run functional tests.
# TODO: Command line API for bash output filanames & install dir, cache.
if download_cache:
assert os.path.isdir(download_cache), download_cache
# Effectively using this as a global variable, refactor this
# once using a visitor pattern instead of action.to_bash()
os.environ["DOWNLOAD_CACHE"] = os.path.abspath(download_cache)
print("Using $DOWNLOAD_CACHE=%r" % os.environ["DOWNLOAD_CACHE"])
failed = False
with open("", "w") as env_sh_handle:
with open("", "w") as install_handle:
for path in paths:
# ctx.log("Checking: %r" % path)
if failed and fail_fast:
for tool_dep in find_tool_dependencis_xml(path, recursive):
passed = process_tool_dependencies_xml(tool_dep,
if passed:
info('Processed %s' % tool_dep)
failed = True
if fail_fast:
for line in [
'#' + '*' * 60,
'echo "WARNING: Skipping %s"' % tool_dep,
'#' + '*' * 60]:
install_handle.write(line + "\n")
# error("%s failed" % tool_dep)
ctx.log("The End")
if failed:
error('Error processing one or more tool_dependencies.xml files.')
@@ -620,3 +620,15 @@ def _compose(*functions):
def compose2(f, g):
return lambda x: f(g(x))
return functools.reduce(compose2, functions)

def dependencies_script_options():
return _compose(
type=click.Path(file_okay=False, resolve_path=True),
help=("Directory to cache downloaded files, default is $DOWNLOAD_CACHE"),
Oops, something went wrong.

5 comments on commit f798c7e


This comment has been minimized.

Copy link

jmchilton replied Oct 7, 2015

Really nice work @peterjc, thanks for continuing to invest this much time and effort in Galaxy and related efforts 😄.


This comment has been minimized.

Copy link
Contributor Author

peterjc replied Oct 8, 2015

Example usage within TravisCI: peterjc/galaxy_blast@a1303d7 and

In this case I am using planemo dependency_script to run two tool_dependencies.xml files to install BLAST+ 2.2.31 and Blast2GO, which avoids duplicating relatively complex install recipes in .travis.yml.

Note that the dependency chain here is being determined manually - for example at this point I am only wanting packages/package_blast_plus_2_2_31/tool_dependencies.xml for the current wrapper tests, and not the older versions of this package.


This comment has been minimized.

Copy link

nsoranzo replied Oct 8, 2015

@peterjc Neat!


This comment has been minimized.

Copy link
Contributor Author

peterjc replied Oct 8, 2015

Second example usage within TravisCI peterjc/galaxy_mira@0d41761 and

In this case planemo dependency_script is used on three tool_dependencies.xml files to install three versions of MIRA. There is a complication of clashing binary names requiring installation into separate folders and determining priority on $PATH (see #319 and #320).


This comment has been minimized.

Copy link
Contributor Author

peterjc replied Oct 22, 2015

Third example usage within TravisCI peterjc/pico_galaxy@6d20c30 and

Much like the previous examples, this uses planemo dependency_script on several tool_dependencies.xml files to install binaries then used to test the tool wrappers.

Fourth example usage within TravisCI biopython/galaxy_packages@91c4d21 and

This example is a little different to the previous ones. Here the repository only contains Galaxy dependency packages, and planemo dependency_script is used essentially for minimal validation of the tool_dependencies.xml files (and to check their download URLs are working).

Please sign in to comment.