[rfr] Credible set visualization options

[abought]

🦑 This is a work in progress and should not yet be merged

Ticket

https://github.com/statgen/locuszoom/issues/39

Purpose

Implement credible sets with new visualizations. This PR exists to track state of work and make visible to others for collaboration. Some user-facing hooks are in flux or need unit tests added.

Summary of changes

• Consolidate build tooling and add sourcemap support for easier debugging. Removes vendor files from repo because unminified are needed for sourcemap creation.

• Update node dev dependencies to newer versions (keep vendor build dependencies pinned_

• Add new "display options" dashboard component for flexible visualization choices (eg color by fields)

• New example page showing use of external credible sets library

TODO

• [x] NPM release of credible sets library; point example page at a tagged release version (with better CDN link)
• [x] Better explanation/ intro docs within demo page (or link to docs for standalone library)
• [x] Unit tests for new layers/api hooks

Known issues

• gulp-wrap does not support proper source maps so app build JS line numbers are consistently off by ~40 lines
• Pre-existing bug: two linked tracks can get out of sync if user pans around a lot: 1) drag a top plot (like association scatter) to the left of the screen about halfway across the range; 2) drag a bottom plot (like genes) halfway across the range to the right; 3) repeat a few times, and eventually the genes won't line up under the same place in the scatter plot.

Testing notes

New functionality

• New annotation track
• Dashboard button for display options. Should be able to control point size, point shape, color, and legend using existing layout directives.

Potential regressions

• The build process has been reworked, and built assets are now deposited in the dist folder (to support direct NPM installs via Git URL)

Release notes

LocusZoom 0.7.0

New features

• Single-item annotation track: customizable track to identify points that meet certain criteria. Useful when annotating specific data points alongside other information like LD: allows rich display of many kinds of information on the same plot.
• Display options dashboard button: interactively apply different colors and styles to
• New demonstration: annotate plots with region-based credible sets based on standalone gwas-credible-sets library

Internal improvements

• Revamped build process; we now provide source maps for vendor files to facilitate debugging
• Built assets have been moved to consolidated dist directory
• Fix issue with tooltips not appearing correctly in interval annotation plot

[abought]

This branch is largely feature complete. Tomorrow will be on various tidying, testing, and user feedback, with a target to package and release by noon Friday at the latest.

This PR adds a new dashboard button + new data layer type. Whether before or after the holidays, any feedback on desired features (or implementation gotchas) would be welcomed. (eg @Frencil etc) The actual calcs are implemented in a separate repo.

Hope everyone has an excellent break!

[abought]

Added some unit tests, and at least for the exact method, results match R output. (haven't manually ported over all the extreme value cases)

@welchr, current goal is to package this up and cut an LZ release by ~12pm tomorrow, based on current state of branch. I'll hold off if you have change requests and/or if we can't resolve @Frencil 's tooling issue report by that time.

Last call for tweaks!

EDIT: I've also added docs for the annotation track and display options to the wiki, since these are user-facing features of possible interest.

[abought]

Hmm. I'm tentatively marking the blocker as resolved, based on a few trials (below). If problems persist, please let me know and we can investigate further.

I've run the build/install process with 6.4, 5.12, 6.12 (LTS-6), and 8.9.3 (LTS-8), on old (6+ months) as well as new copies of the repo and on a different computer. I also tried with wiping node_modules after each install (to ensure bad dependencies weren't getting "grandfathered in").

The one issue I saw could be solved by good old npm rebuild node-sass --force; after that assets could be built normally on all machines/versions.