2005
92%
master: 89%

Ran 11 Nov 2014 12:54PM UTC

Jobs 1

Files 52

Run time 22s

Badge

Embed ▾

pending completion

Build # 2005

Build Type

push

travis-ci

Committed by

tombooth

Commit Message

Improve perf characteristics of filling gaps

Previously for every period and expected permutation of group_by key
value it would loop over all data points trying to find a match. This
led to num_periods * num_permutations * num_data_points iterations,
which will get very slow very fast. We will always need to go through
every period and permutation but removing the last multiple has big
change.

I switched this for pre-hashing the data points so that we can find a
match in O(1) rather than O(num_data_points). The reduced the time to
fill in gaps from 180s to 0.175s for the troublesome lpa journey query.

I have ensured that the output of this function is the same. The pickled
arguments and results are pulled out by wrapping
fill_group_by_permutations in a decorator that grabs *args and **kwargs
and the results and stores them in a file.

You can see the result of the profile by running:
	`nosetests --nocapture tests/perf/`

Run Details

1797 of 1961 relevant lines covered (91.64%)

0.92 hits per line

Jobs

ID	Job ID	Ran	Files	Coverage
1	2005.1 (SKIP_VIRUS_SCAN=1 SKIP_SPLINTER_TESTS=1 MONGO_REPLICA_SET='' NO_AUTOPEP8=1)	11 Nov 2014 12:54PM UTC	0	91.64	Travis Job 2005.1

Source Files on build 2005

Detailed source file information is not available for this build.

alphagov / backdrop / 2005
92%
master: 89%

README BADGES
x

Markdown

Textile

RDoc

HTML

Rst

Jobs

Source Files on build 2005

alphagov / backdrop / 2005 92% master: 89%

README BADGES x

Markdown

Textile

RDoc

HTML

Rst

Jobs

Source Files on build 2005

alphagov / backdrop / 2005
92%
master: 89%

README BADGES
x