README.md
1pipeline
2========
3
4This directory contains tools and scripts for running a cron job that does
5RAPPOR analysis and generates an HTML dashboard.
6
7It works like this:
8
91. `task_spec.py` generates a text file where each line corresponds to a process
10 to be run (a "task"). The process is `bin/decode-dist` or
11 `bin/decode-assoc`. The line contains the task parameters.
12
132. `xargs -P` is used to run processes in parallel. Our analysis is generally
14 single-threaded (i.e. because R is single-threaded), so this helps utilize
15 the machine fully. Each task places its output in a different subdirectory.
16
173. `cook.sh` calls `combine_results.py` to combine analysis results into a time
18 series. It also calls `combine_status.py` to keep track of task data for
19 "meta-analysis". `metric_status.R` generates more summary CSV files.
20
214. `ui.sh` calls `csv_to_html.py` to generate an HTML fragments from the CSV
22 files.
23
245. The JavaScript in `ui/ui.js` is loaded from static HTML, and makes AJAX calls
25 to retrieve the HTML fragments. The page is made interactive with
26 `ui/table-lib.js`.
27
28`dist.sh` and `assoc.sh` contain functions which coordinate this process.
29
30`alarm-lib.sh` is used to kill processes that have been running for too long.
31
32Testing
33-------
34
35`pipeline/regtest.sh` contains end-to-end demos of this process. Right now it
36depends on testdata from elsewhere in the tree:
37
38
39 rappor$ ./demo.sh run # prepare dist testdata
40 rappor$ cd bin
41
42 bin$ ./test.sh write-assoc-testdata # prepare assoc testdata
43 bin$ cd ../pipeline
44
45 pipeline$ ./regtest.sh dist
46 pipeline$ ./regtest.sh assoc
47
48 pipeline$ python -m SimpleHTTPServer # start a static web server
49
50 http://localhost:8000/_tmp/
51
52
53