comparison README.txt @ 39:64f7bd086146 draft

Uploaded
author fubar
date Thu, 28 Aug 2014 03:05:01 -0400
parents 9b64aa8688ad
children 8603867451d8
comparison
equal deleted inserted replaced
38:5bf5c411baf3 39:64f7bd086146
1 # WARNING before you start
2 # Install this tool on a private Galaxy ONLY
3 # Please NEVER on a public or production instance
4 # updated august 8 2014 to fix bugs reported by Marius van den Beek
5 Please cite:
6 http://bioinformatics.oxfordjournals.org/cgi/reprint/bts573?ijkey=lczQh1sWrMwdYWJ&keytype=ref
7 if you use this tool in your published work.
8
9 *Short Story*
10
11 This is an unusual Galaxy tool that exposes unrestricted and therefore extremely dangerous
12 scripting to designated administrative users of a Galaxy server, allowing them to run scripts
13 in R, python, sh and perl over a single input data set, writing a single new data set as output.
14
15 In addition, this tool optionally generates very simple new Galaxy tools, that effectively
16 freeze the supplied script into a new, ordinary Galaxy tool that runs it over a single input file,
17 working just like any other Galaxy tool for your users.
18
19 If you use the Html output option, please ensure that sanitize_all_html is set to False and
20 uncommented in universe_wsgi.ini - it should show:
21
22 # By default, all tool output served as 'text/html' will be sanitized
23 sanitize_all_html = False
24
25 *More Detail*
26
27 To use the ToolFactory, you should have prepared a script to paste into a text box,
28 and a small test input example ready to select from your history to test your new script.
29 There is an example in each scripting language on the Tool Factory form. You can just
30 cut and paste these to try it out - remember to select the right interpreter please. You'll
31 also need to create a small test data set using the Galaxy history add new data tool.
32
33 If the script fails somehow, use the "redo" button on the tool output in your history to
34 recreate the form complete with broken script. Fix the bug and execute again. Rinse, wash, repeat.
35
36 Once the script runs sucessfully, a new Galaxy tool that runs your script can be generated.
37 Select the "generate" option and supply some help text and names. The new tool will be
38 generated in the form of a new Galaxy datatype - toolshed.gz - as the name suggests,
39 it's an archive ready to upload to a Galaxy ToolShed as a new tool repository.
40
41 Once it's in a ToolShed, it can be installed into any local Galaxy server from
42 the server administrative interface.
43
44 Once the new tool is installed, local users can run it - each time, the script that was supplied
45 when it was built will be executed with the input chosen from the user's history. In other words,
46 the tools you generate with the ToolFactory run just like any other Galaxy tool,
47 but run your script every time.
48
49 Tool factory tools are perfect for workflow components. One input, one output, no variables.
50
51 *To fully and safely exploit the awesome power* of this tool, Galaxy and the ToolShed,
52 you should be a developer installing this tool on a private/personal/scratch local instance where you
53 are an admin_user. Then, if you break it, you get to keep all the pieces
54 see https://bitbucket.org/fubar/galaxytoolfactory/wiki/Home
55
56 ** Installation **
57 This is a Galaxy tool. You can install it most conveniently using the administrative "Search and browse tool sheds" link.
58 Find the Galaxy Main toolshed at https://toolshed.g2.bx.psu.edu/ and search for the toolfactory repository.
59 Open it and review the code and select the option to install it.
60
61 (
62 If you can't get the tool that way, the xml and py files here need to be copied into a new tools
63 subdirectory such as tools/toolfactory Your tool_conf.xml needs a new entry pointing to the xml
64 file - something like::
65
66 <section name="Tool building tools" id="toolbuilders">
67 <tool file="toolfactory/rgToolFactory.xml"/>
68 </section>
69
70 If not already there (I just added it to datatypes_conf.xml.sample), please add:
71 <datatype extension="toolshed.gz" type="galaxy.datatypes.binary:Binary" mimetype="multipart/x-gzip" subclass="True" />
72 to your local data_types_conf.xml.
73 )
74
75 Of course, R, python, perl etc are needed on your path if you want to test scripts using those interpreters.
76 Adding new ones to this tool code should be easy enough. Please make suggestions as bitbucket issues and code.
77 The HTML file code automatically shrinks R's bloated pdfs, and depends on ghostscript. The thumbnails require imagemagick .
78
79 * Restricted execution *
80 The new tool factory tool will then be usable ONLY by admin users - people with IDs in admin_users in universe_wsgi.ini
81 **Yes, that's right. ONLY admin_users can run this tool** Think about it for a moment. If allowed to run any
82 arbitrary script on your Galaxy server, the only thing that would impede a miscreant bent on destroying all your
83 Galaxy data would probably be lack of appropriate technical skills.
84
85 *What it does* This is a tool factory for simple scripts in python, R and perl currently.
86 Functional tests are automatically generated. How cool is that.
87
88 LIMITED to simple scripts that read one input from the history.
89 Optionally can write one new history dataset,
90 and optionally collect any number of outputs into links on an autogenerated HTML
91 index page for the user to navigate - useful if the script writes images and output files - pdf outputs
92 are shown as thumbnails and R's bloated pdf's are shrunk with ghostscript so that and imagemagik need to
93 be avaailable.
94
95 Generated tools can be edited and enhanced like any Galaxy tool, so start small and build up since
96 a generated script gets you a serious leg up to a more complex one.
97
98 *What you do* You paste and run your script
99 you fix the syntax errors and eventually it runs
100 You can use the redo button and edit the script before
101 trying to rerun it as you debug - it works pretty well.
102
103 Once the script works on some test data, you can
104 generate a toolshed compatible gzip file
105 containing your script ready to run as an ordinary Galaxy tool in a
106 repository on your local toolshed. That means safe and largely automated installation in any
107 production Galaxy configured to use your toolshed.
108
109 *Generated tool Security* Once you install a generated tool, it's just
110 another tool - assuming the script is safe. They just run normally and their user cannot do anything unusually insecure
111 but please, practice safe toolshed.
112 Read the fucking code before you install any tool.
113 Especially this one - it is really scary.
114
115 If you opt for an HTML output, you get all the script outputs arranged
116 as a single Html history item - all output files are linked, thumbnails for all the pdfs.
117 Ugly but really inexpensive.
118
119 Patches and suggestions welcome as bitbucket issues please?
120
121 copyright ross lazarus (ross stop lazarus at gmail stop com) May 2012
122
123 all rights reserved
124 Licensed under the LGPL if you want to improve it, feel free https://bitbucket.org/fubar/galaxytoolfactory/wiki/Home
125
126 Material for our more enthusiastic and voracious readers continues below - we salute you.
127
128 **Motivation** Simple transformation, filtering or reporting scripts get written, run and lost every day in most busy labs
129 - even ours where Galaxy is in use. This 'dark script matter' is pervasive and generally not reproducible.
130
131 **Benefits** For our group, this allows Galaxy to fill that important dark script gap - all those "small" bioinformatics
132 tasks. Once a user has a working R (or python or perl) script that does something Galaxy cannot currently do (eg transpose a
133 tabular file) and takes parameters the way Galaxy supplies them (see example below), they:
134
135 1. Install the tool factory on a personal private instance
136
137 2. Upload a small test data set
138
139 3. Paste the script into the 'script' text box and iteratively run the insecure tool on test data until it works right -
140 there is absolutely no reason to do this anywhere other than on a personal private instance.
141
142 4. Once it works right, set the 'Generate toolshed gzip' option and run it again.
143
144 5. A toolshed style gzip appears ready to upload and install like any other Toolshed entry.
145
146 6. Upload the new tool to the toolshed
147
148 7. Ask the local admin to check the new tool to confirm it's not evil and install it in the local production galaxy
149
150 **Simple examples on the tool form**
151
152 A simple Rscript "filter" showing how the command line parameters can be handled, takes an input file,
153 does something (transpose in this case) and writes the results to a new tabular file::
154
155 # transpose a tabular input file and write as a tabular output file
156 ourargs = commandArgs(TRUE)
157 inf = ourargs[1]
158 outf = ourargs[2]
159 inp = read.table(inf,head=F,row.names=NULL,sep='\t')
160 outp = t(inp)
161 write.table(outp,outf, quote=FALSE, sep="\t",row.names=F,col.names=F)
162
163 Calculate a multiple test adjusted p value from a column of p values - for this script to be useful,
164 it needs the right column for the input to be specified in the code for the
165 given input file type(s) specified when the tool is generated ::
166
167 # use p.adjust - assumes a HEADER row and column 1 - please fix for any real use
168 column = 1 # adjust if necessary for some other kind of input
169 fdrmeth = 'BH'
170 ourargs = commandArgs(TRUE)
171 inf = ourargs[1]
172 outf = ourargs[2]
173 inp = read.table(inf,head=T,row.names=NULL,sep='\t')
174 p = inp[,column]
175 q = p.adjust(p,method=fdrmeth)
176 newval = paste(fdrmeth,'p-value',sep='_')
177 q = data.frame(q)
178 names(q) = newval
179 outp = cbind(inp,newval=q)
180 write.table(outp,outf, quote=FALSE, sep="\t",row.names=F,col.names=T)
181
182
183
184 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
185 turned on for this to work. The heatmap will be presented as a thumbnail linked to the pdf in the resulting HTML page::
186
187 # note this script takes NO input or output because it generates random data
188 foo = data.frame(a=runif(100),b=runif(100),c=runif(100),d=runif(100),e=runif(100),f=runif(100))
189 bar = as.matrix(foo)
190 pdf( "heattest.pdf" )
191 heatmap(bar,main='Random Heatmap')
192 dev.off()
193
194 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
195 and pasted into the script box. Note that you can already do this in Galaxy by setting up the cut columns tool with the
196 correct number of columns in reverse order,but this script will work for any number of columns so is completely generic::
197
198 # reverse order of columns in a tabular file
199 import sys
200 inp = sys.argv[1]
201 outp = sys.argv[2]
202 i = open(inp,'r')
203 o = open(outp,'w')
204 for row in i:
205 rs = row.rstrip().split('\t')
206 rs.reverse()
207 o.write('\t'.join(rs))
208 o.write('\n')
209 i.close()
210 o.close()
211
212
213 Galaxy as an IDE for developing API scripts
214 If you need to develop Galaxy API scripts and you like to live dangerously, please read on.
215
216 Galaxy as an IDE?
217 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.
218
219 Why bother - what's wrong with Eclipse
220 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!
221
222 Workflow
223 Fire up the Tool Factory in Galaxy.
224
225 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.
226
227 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.
228
229 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.
230
231 Eg tool factory api script
232 import sys
233 from blend.galaxy import GalaxyInstance
234 ourGal = 'http://x.x.x.x:xxxx'
235 ourKey = 'xxx'
236 gi = GalaxyInstance(ourGal, key=ourKey)
237 libs = gi.libraries.get_libraries()
238 res = []
239 # libs looks like
240 # u'url': u'/galaxy/api/libraries/441d8112651dc2f3', u'id': u'441d8112651dc2f3', u'name':.... u'Demonstration sample RNA data',
241 for lib in libs:
242 res.append('%s:\n' % lib['name'])
243 res.append(str(gi.libraries.show_library(lib['id'],contents=True)))
244 outf=open(sys.argv[2],'w')
245 outf.write('\n'.join(res))
246 outf.close()
247
248 **Attribution**
249 Creating re-usable tools from scripts: The Galaxy Tool Factory
250 Ross Lazarus; Antony Kaspi; Mark Ziemann; The Galaxy Team
251 Bioinformatics 2012; doi: 10.1093/bioinformatics/bts573
252
253 http://bioinformatics.oxfordjournals.org/cgi/reprint/bts573?ijkey=lczQh1sWrMwdYWJ&keytype=ref
254
255 **Licensing**
256 Copyright Ross Lazarus 2010
257 ross lazarus at g mail period com
258
259 All rights reserved.
260
261 Licensed under the LGPL
262
263 **Obligatory screenshot**
264
265 http://bitbucket.org/fubar/galaxytoolmaker/src/fda8032fe989/images/dynamicScriptTool.png
266