Perfomatic:UI: Difference between revisions
Jump to navigation
Jump to search
(tools) |
(linewidth) |
||
| Line 45: | Line 45: | ||
* <strike>add caching proxy (varnish, etc)</strike> | * <strike>add caching proxy (varnish, etc)</strike> | ||
* improve charting performance | * improve charting performance | ||
** forcing lineWidth to 1 seems to make a huge difference | |||
*** started [https://groups.google.com/forum/#!topic/flot-graphs/tJDCYMe-0Nw discussion on the flot list] | |||
*** however does not look as good, not pushing to stage until further investigation | |||
** "double-buffering" canvas for flot | ** "double-buffering" canvas for flot | ||
*** [http://stackoverflow.com/questions/2795269/does-html5-canvas-support-double-buffering/2864533#2864533 stack overflow discussion] | *** [http://stackoverflow.com/questions/2795269/does-html5-canvas-support-double-buffering/2864533#2864533 stack overflow discussion] | ||
Revision as of 19:38, 1 January 2011
Design
Designed by Chris Howse
Implementation
Missing features from current design
- dashboard ->
- toolbar ->
product - not implementedplatforms - not implementedtop tests - not implementedset pulldowns based on URL
- data grid ->
- row/column should not scroll
- toolbar ->
- custom chart ->
- toolbar ->
running time - time delta/time delta % not implementedshow changesets - needs design workmay need hgpushlog work to link directly to changesets
- export CSV - server broken
- link - needs url shortener
url/embed styling prototype changes
embed - not implemented
set selection range based on URL- add test data ->
- put popular branches/tests/platforms on top
- toolbar ->
- plot ->
y-axis should always be 0-based
need mapping between branches and hg repos
Performance optimizations
- split up JSON files as discussed in status meeting
- look into constraining by date range
set Last-Modified headers on all JSON feedsespecially the manifest!already setting 15m Expires/Cache-control max-age
add caching proxy (varnish, etc)- improve charting performance
- forcing lineWidth to 1 seems to make a huge difference
- started discussion on the flot list
- however does not look as good, not pushing to stage until further investigation
- "double-buffering" canvas for flot
- stack overflow discussion
- custom chart is much faster if each series is written to a hidden canvas then shown when they are all complete
use node.js to generate static images for dashboard?flot group post- this actually works pretty well
- spoke to author (John Schulz, jsno.de/) who linked the code behind http://jsno.de./flot/
requires a text plugin for flot, to get the ticks into the canvas instead of flot's default of appending styled divs
- look into reducing number of points
- flot plots way more than is actually useful when graphs are very dense, flot group discussion
- could filter out points on the server side, using e.g. k-means clustering
- look into flot alternatives
- forcing lineWidth to 1 seems to make a huge difference
Testing
- stand up vhost with old graphserver, pointing at nightly database, for QA comparison
- set up lint and tests on Hudson
- selenium tests (in progress)
- show these to QA to run on their Hudson
- qUnit/env.js unit tests for JS code
- selenium tests (in progress)
- stand up auto-update for code (5 minutes) and DB (nightly)
- (QA) compare actual submitted data to JSON, and confirm in interface
- make sure proper changeset range is shown for checkins
- check with catlee on how the tree-management script does this
- follow-up irc convo w/ ted on catlee on changesets coalescing