view README.txt @ 3:d75fd880fce2 draft

author fubar
date Sun, 11 Jan 2015 22:01:27 -0500
parents c34063ab3735
children dd6cf2ddaac7
line wrap: on
line source

# WARNING before you start
# Install this tool on a private Galaxy ONLY
# Please NEVER on a public or production instance
# updated august 2014 by John Chilton adding citation support
# updated august 8 2014 to fix bugs reported by Marius van den Beek
# please cite the resource at 
# if you use this tool in your published work.

*Short Story*

This is an unusual Galaxy tool capable of generating new Galaxy tools.
It works by exposing *unrestricted* and therefore extremely dangerous
scripting to all designated administrators of the host Galaxy server, allowing them to run scripts 
in R, python, sh and perl over multiple selected input data sets, writing a single new data set as output.

*Automated outputs in named sections*

If your script writes to the current directory path, arbitrary mix of (eg) pdfs, tabular analysis results and run logs,
the tool factory can optionally auto-generate a linked Html page with separate sections showing a thumbnail grid
for all pdfs and the log text, grouping all artifacts sharing a file name and log name prefix::

 eg: if "foo.log" is emitted then *all* other outputs matching foo_* will all be grouped together - eg
 foo_bar.pdf and
 would all be displayed and linked in the same section with foo.log's contents - to form the "Foo" section of the Html page.
 Sections appear in alphabetic order and there are no limits on the number of files or sections.

*Automated generation of new Galaxy tool shed tools for installation into any Galaxy*

Once a script is working correctly, this tool optionally generates a new Galaxy tool, effectively
freezing the supplied script into a new, ordinary Galaxy tool that runs it over one or more input files
selected by the user. Generated tools are installed via a tool shed by an administrator and work exactly like all other Galaxy tools for your users. 

If you use the Html output option, please ensure that sanitize_all_html is set to False and 
uncommented in universe_wsgi.ini - it should show::

 # By default, all tool output served as 'text/html' will be sanitized
 sanitize_all_html = False

This opens potential security risks and may not be acceptable for public sites where the lack of stylesheets
may make Html pages damage onlookers' eyeballs but should still be correct.

*More Detail*

To use the ToolFactory, you should have prepared a script to paste into a text box,
and a small test input example ready to select from your history to test your new script.
There is an example in each scripting language on the Tool Factory form. You can just
cut and paste these to try it out - remember to select the right interpreter please. You'll
also need to create a small test data set using the Galaxy history add new data tool.

If the script fails somehow, use the "redo" button on the tool output in your history to
recreate the form complete with broken script. Fix the bug and execute again. Rinse, wash, repeat.

Once the script runs sucessfully, a new Galaxy tool that runs your script can be generated.
Select the "generate" option and supply some help text and names. The new tool will be
generated in the form of a new Galaxy datatype - toolshed.gz - as the name suggests,
it's an archive ready to upload to a Galaxy ToolShed as a new tool repository.

Once it's in a ToolShed, it can be installed into any local Galaxy server from
the server administrative interface.

Once the new tool is installed, local users can run it - each time, the script that was supplied
when it was built will be executed with the input chosen from the user's history. In other words,
the tools you generate with the ToolFactory run just like any other Galaxy tool,
but run your script every time. 

Tool factory tools are perfect for workflow components. One input, one output, no variables.

*To fully and safely exploit the awesome power* of this tool, Galaxy and the ToolShed,
you should be a developer installing this tool on a private/personal/scratch local instance where you 
are an admin_user. Then, if you break it, you get to keep all the pieces

** Installation **
This is a Galaxy tool. You can install it most conveniently using the administrative "Search and browse tool sheds" link.
Find the Galaxy Main toolshed at and search for the toolfactory repository.
Open it and review the code and select the option to install it.

If you can't get the tool that way, the xml and py files here need to be copied into a new tools 
subdirectory such as tools/toolfactory Your tool_conf.xml needs a new entry pointing to the xml 
file - something like::

  <section name="Tool building tools" id="toolbuilders">
    <tool file="toolfactory/rgToolFactory.xml"/>

If not already there (I just added it to datatypes_conf.xml.sample), please add:
<datatype extension="toolshed.gz" type="galaxy.datatypes.binary:Binary" mimetype="multipart/x-gzip" subclass="True" />
to your local data_types_conf.xml.

Of course, R, python, perl etc are needed on your path if you want to test scripts using those interpreters.
Adding new ones to this tool code should be easy enough. Please make suggestions as bitbucket issues and code.
The HTML file code automatically shrinks R's bloated pdfs, and depends on ghostscript. The thumbnails require imagemagick .

* Restricted execution *
The tool factory tool itself will then be usable ONLY by admin users - people with IDs in admin_users in universe_wsgi.ini
**Yes, that's right. ONLY admin_users can run this tool** Think about it for a moment. If allowed to run any
arbitrary script on your Galaxy server, the only thing that would impede a miscreant bent on destroying all your
Galaxy data would probably be lack of appropriate technical skills.

*What it does* This is a tool factory for simple scripts in python, R and perl currently. 
Functional tests are automatically generated. How cool is that. 

LIMITED to simple scripts that read one input from the history.
Optionally can write one new history dataset,
and optionally collect any number of outputs into links on an autogenerated HTML
index page for the user to navigate - useful if the script writes images and output files - pdf outputs
are shown as thumbnails and R's bloated pdf's are shrunk with ghostscript so that and imagemagik need to
be avaailable.

Generated tools can be edited and enhanced like any Galaxy tool, so start small and build up since
a generated script gets you a serious leg up to a more complex one.

*What you do* You paste and run your script
you fix the syntax errors and eventually it runs
You can use the redo button and edit the script before
trying to rerun it as you debug - it works pretty well.

Once the script works on some test data, you can
generate a toolshed compatible gzip file
containing your script ready to run as an ordinary Galaxy tool in a
repository on your local toolshed. That means safe and largely automated installation in any
production Galaxy configured to use your toolshed.

*Generated tool Security* Once you install a generated tool, it's just
another tool - assuming the script is safe. They just run normally and their user cannot do anything unusually insecure
but please, practice safe toolshed.
Read the fucking code before you install any tool. 
Especially this one - it is really scary.

If you opt for an HTML output, you get all the script outputs arranged
as a single Html history item - all output files are linked, thumbnails for all the pdfs.
Ugly but really inexpensive.

Patches and suggestions welcome as bitbucket issues please? 

copyright ross lazarus (ross stop lazarus at gmail stop com) May 2012

all rights reserved
Licensed under the LGPL if you want to improve it, feel free

Material for our more enthusiastic and voracious readers continues below - we salute you.

**Motivation** Simple transformation, filtering or reporting scripts get written, run and lost every day in most busy labs 
- even ours where Galaxy is in use. This 'dark script matter' is pervasive and generally not reproducible.

**Benefits** For our group, this allows Galaxy to fill that important dark script gap - all those "small" bioinformatics 
tasks. Once a user has a working R (or python or perl) script that does something Galaxy cannot currently do (eg transpose a 
tabular file) and takes parameters the way Galaxy supplies them (see example below), they:

1. Install the tool factory on a personal private instance

2. Upload a small test data set

3. Paste the script into the 'script' text box and iteratively run the insecure tool on test data until it works right - 
there is absolutely no reason to do this anywhere other than on a personal private instance. 

4. Once it works right, set the 'Generate toolshed gzip' option and run it again. 

5. A toolshed style gzip appears ready to upload and install like any other Toolshed entry. 

6. Upload the new tool to the toolshed

7. Ask the local admin to check the new tool to confirm it's not evil and install it in the local production galaxy

**Simple examples on the tool form**

A simple Rscript "filter" showing how the command line parameters can be handled, takes an input file, 
does something (transpose in this case) and writes the results to a new tabular file::

 # transpose a tabular input file and write as a tabular output file
 ourargs = commandArgs(TRUE)
 inf = ourargs[1]
 outf = ourargs[2]
 inp = read.table(inf,head=F,row.names=NULL,sep='\t')
 outp = t(inp)
 write.table(outp,outf, quote=FALSE, sep="\t",row.names=F,col.names=F)

Calculate a multiple test adjusted p value from a column of p values - for this script to be useful,
it needs the right column for the input to be specified in the code for the
given input file type(s) specified when the tool is generated ::

 # use p.adjust - assumes a HEADER row and column 1 - please fix for any real use
 column = 1 # adjust if necessary for some other kind of input
 fdrmeth = 'BH'
 ourargs = commandArgs(TRUE)
 inf = ourargs[1]
 outf = ourargs[2]
 inp = read.table(inf,head=T,row.names=NULL,sep='\t')
 p = inp[,column]
 q = p.adjust(p,method=fdrmeth)
 newval = paste(fdrmeth,'p-value',sep='_')
 q = data.frame(q)
 names(q) = newval
 outp = cbind(inp,newval=q)
 write.table(outp,outf, quote=FALSE, sep="\t",row.names=F,col.names=T) 

Another Rscript example without any input file - generates a random heatmap pdf - you must make sure the option to create an HTML output file is
turned on for this to work. The heatmap will be presented as a thumbnail linked to the pdf in the resulting HTML page::

 # note this script takes NO input or output because it generates random data
 foo = data.frame(a=runif(100),b=runif(100),c=runif(100),d=runif(100),e=runif(100),f=runif(100))
 bar = as.matrix(foo)
 pdf( "heattest.pdf" )
 heatmap(bar,main='Random Heatmap')

A Python example that reverses each row of a tabular file. You'll need to remove the leading spaces for this to work if cut
and pasted into the script box. Note that you can already do this in Galaxy by setting up the cut columns tool with the
correct number of columns in reverse order,but this script will work for any number of columns so is completely generic::

# reverse order of columns in a tabular file
import sys
inp = sys.argv[1]
outp = sys.argv[2]
i = open(inp,'r')
o = open(outp,'w')
for row in i:
    rs = row.rstrip().split('\t')

Galaxy as an IDE for developing API scripts
If you need to develop Galaxy API scripts and you like to live dangerously, please read on.

Galaxy as an IDE?
Amazingly enough, blend-lib API scripts run perfectly well *inside* Galaxy when pasted into a Tool Factory form. No need to generate a new tool. Galaxy+Tool_Factory = IDE I think we need a new t-shirt. Seriously, it is actually quite useable.

Why bother - what's wrong with Eclipse
Nothing. But, compared with developing API scripts in the usual way outside Galaxy, you get persistence and other framework benefits plus at absolutely no extra charge, a ginormous security problem if you share the history or any outputs because they contain the api script with key so development servers only please!

Fire up the Tool Factory in Galaxy.

Leave the input box empty, set the interpreter to python, paste and run an api script - eg working example (substitute the url and key) below.

It took me a few iterations to develop the example below because I know almost nothing about the API. I started with very simple code from one of the samples and after each run, the (edited..) api script is conveniently recreated using the redo button on the history output item. So each successive version of the developing api script you run is persisted - ready to be edited and rerun easily. It is ''very'' handy to be able to add a line of code to the script and run it, then view the output to (eg) inspect dicts returned by API calls to help move progressively deeper iteratively.

Give the below a whirl on a private clone (install the tool factory from the main toolshed) and try adding complexity with few rerun/edit/rerun cycles.

Eg tool factory api script
import sys
from blend.galaxy import GalaxyInstance
ourGal = 'http://x.x.x.x:xxxx'
ourKey = 'xxx'
gi = GalaxyInstance(ourGal, key=ourKey)
libs = gi.libraries.get_libraries()
res = []
# libs looks like
# u'url': u'/galaxy/api/libraries/441d8112651dc2f3', u'id': u'441d8112651dc2f3', u'name':.... u'Demonstration sample RNA data',
for lib in libs:  
    res.append('%s:\n' % lib['name'])

Creating re-usable tools from scripts: The Galaxy Tool Factory 
Ross Lazarus; Antony Kaspi; Mark Ziemann; The Galaxy Team
Bioinformatics 2012; doi: 10.1093/bioinformatics/bts573

Copyright Ross Lazarus 2010
ross lazarus at g mail period com

All rights reserved.

Licensed under the LGPL

**Obligatory screenshot**