bwinton / australistelemetry Goto Github PK

Week number? Dates? Dates and version? Dates and version and channel?

Given the channel and versions, do we even need the fields buildID and submission_date, or are they redundant? We can add these fields as optional arguments for the very delicate analysis later, but they don't seem to be necessary right now with the other fields specified (correct me if i'm wrong!!!) Should we just set these from 0 to 999999 by default?

direct failures somewhere other than console.

proc.communicate() - why isn't it returning anything?

Let user feed in a "tag" at command line (ie, week 43 aurora) to label all output including filters.

ex: channel

as in: http://www.artima.com/weblogs/viewpost.jsp?thread=4829

Especially individual steps at the bottom of australis_report_gen

• single user actions in aggregate; ex: click on more than one kind of search?

"%03d" % (7,)
10:52 "%02d" % (7,)
10:53 In [3]: "%03d" % (7,)
10:53 Out[3]: '007'
10:54 In [3]: "%03d-%s" % (7,tag)
10:54 current!

Which? What are their defaults?
Specifically:
args = {

• "job_script" : "../uitour.py",
• "input_filter": filter,
• "num_mappers" : 16,
• "num_reducers" : 4,
• "data_dir" : "../work/cache",
• "work_dir" : "../work",
• "output" : output_file,
• "bucket" : "telemetry-published-v1",
• "local_only" : True
• }

many solutions:
-make a job2.py that takes a dict config (gross - won't get updates)
-feed in a namedtuple to original job to keep the dot notation (also gross - hard to update/read code and bloat)
-require everyone the use the new job.py and take in a dict or **kwargs (grossest)

right now, assume 2014 for simplicity

open up pref?

for now, only use x.0 - x.999

-throw error if multiple channels and current version (what version do we mean??)
-if unspecified, take current version, throw error if not
--> find current version from http://bwinton.github.io/releaseboard/example/
if no --version data, give all versions
--week=current give last full week

old countable_events fails