ExonMostlyInitialDesignMeetingWhiteboard

left whiteboard

right whiteboard

Transcription

Please do correct any errors and feel free to clarify things but please keep major discussions on ExonMostlyInitialDesignMeeting or a subpage.

Jim's sections

Browser steps

• Load cart
• Get groups and tracks (Angie sidenote: {groups, supertracks/composites/etc, and user groupings})
• Load data from some displayed tracks
• Calculate track heights & layout
• Render tracks including[?] in loop (Angie sidenote: consider moving series of rendering loops into each track)
  • loading data for rmsk[?], bigWig
  • drawing data
  • making image map for Javascript
• Interpreting Javascript

Multiple regions in zooming and scrolling

• Need some sort of global idea of genomic regions
• algorithm or list

• Fetch some items overlapping a range
• Merge some items that span multiple ranges
• Render item across all ranges

Packing

• Allocates low res array for a line
• Allocates new lines as needed
• Input: list of horizontal regions

Genome regions we support

• region in view
• items in genome coordinates
• items in pixel coordinates

Angie's sections

Incarnations of region list

• regions = {chr1:100-200, chr1:150-250, chr1:500-600, chr2:300-400}

 (transform to displayed regions: merge overlap, add adding)

• dRegions = {chr1:95-255, chr1:495-605, chr2:295-405}

 (transform to fetched regions: merge nearby regions to reduce #db queries)

• fRegions = {chr1:95-605, chr2:295-405}

• dRegions will be mapped to pixel coord ranges

[Emancipate us from series of rendering loops that have to recalculate height exactly in step]

• per-item rendering commands stored in list [by virtual gfx layer]
• quick pass on list calculates total height to allocate for image, packs items[never mind, packing happens at draw time]
• second pass blts into allocated image

Experiments

[partial list]

• how many simultaneous hgTracks requests/processes are typical on the RR?
• how many of the following can be done simultaneously without dragging down system performance?

 * mysql queries
 * processes
 * pthreads

• when is it faster to merge nearby regions for mysql query vs. make multiple queries? [i.e. what is the max distance that we should merge -> dRegions]
• overhead of exec --> socket for parallel mysql requests
• overhead of virtual command list generated by virtual gfx layer