nextstrain / auspice Goto Github PK

Lower right on tree. Move map so that it's in the lower right when this is done.

I'm not sure if this is still a problem or not, but we should revisit now that we're structuring data fetching. This is the only reason we have routes in the dev server - to proxy resources to avoid CORS. In the dev server, the solution to #4 will look something like:

But we could also construct this on the client and request to s3 directly if we solve the CORS issue, which will work in both dev and production. It will require getting localhost:4000 and nextstrain.org requests directly to s3 working.

We should have a config JSON that gives all the options for which datasets can be loaded. This will look something like:

{
"viruses": {
  "flu": {
    "lineage": {
      "H3N2": {
        "duration": {
        "3y",
        "6y",
        "12y" }
      },
     "H1N1pdm": {
        "duration": {
        "3y",
        "6y",
        "12y" }
      }      
    }
  },
  "zika": {}
  }
}

This should be used to create a dataset file string. If this string changes then data should be refetched.

Whenever query params come in, go through this JSON and build dataset.

What?

If you move the bottom handle on the date slider, on http://nextstrain.org/zika/, it removes nodes. Move it back, and not all of them return.

treeplot.call(virusTooltip) invocation seems to be at issue. Within the d3.tip library, el is null. To investigate.

Include multiple tree layout options.

• Rooted tree
  • as mutation count vs Y as layout (current default)
  • as time vs Y as layout (timetree)
  • as time vs Y as mutation count (Path-O-Gen plot)
  • as mutation count vs Y as antigenic distance
• Unrooted tree
• Radial tree

I think this is one dropdown with multiple layout options.

Zoom to selection - put viewport on bounding box of existing nodes

https://github.com/henrythasler/Leaflet.Geodesic

The nextstrain app route to dataset pipeline will look like this:

nextstrain.org/:virus/:strain/:timeperiod, thus: nextstrain.org/flu/h3n2/3y

How we map this to a REST route on the server, and then to string names on s3 is an open question

Each virus (tip) should have an "item page". Each virus has attributes:

• Strain
• Genome string
• Collection date
• Collection location
• Accessions to link out to
• HI assay data
• Multiple sequence alignment with reference viruses
• Tree context
• Papers associated with the virus

Zoom level should adjust to number of nodes on tree at outset so that the appropriate sizes are shown by default

search the node in the tree with the closest sequence. In general, this requires pairwise alignment. I implemented a banded pairwise alignment here:
https://github.com/neherlab/timetree_viewer/blob/master/js/alignPairwise.js

not tested extensively but in contrast to our current version, this should be able to handle gaps.

Richard suggests: "either show but grey, or hide as soon as the date range doesn't include everything."

Currently it is:

And this may be a different decision for different kinds of filters.

Done by auger as a similar fashion to controls, updates list that is provided by auger.

Have a route with path="changelog" that will be hit first and then everything else will fall back on splat.

@trvrb will update meta.json with panels to display

Looks like we need to either:

1. Reroute from error pages via CloudFront. This has the disadvantage of not being indexed by seach engines, so bad.
2. Redirect through an intermediary, like nextstrain.org/zika/ -> nextstrain.org/!/zika/ -> nextstrain.org/zika/, so that S3 puts a /!/ before the path and serves index.html. Not a bad solution. Has extra URL bar action and further hijacks the back button, which isn't good.
3. Drop a bunch of index.html files in the S3 bucket ebola/index.html, zika/index.html, etc... This will result in clean behavior I believe. Just slightly messy in the code base.
4. Spin up a server somewhere.

Some discussion here: http://stackoverflow.com/questions/16267339/s3-static-website-hosting-route-all-paths-to-index-html

d3.json(path + file_prefix + "frequencies.json", function(error, json){
  console.log(error);
  frequencies = json;
  pivots = frequencies["mutations"]["global"]["pivots"].map(function (d) {
    return Math.round(parseFloat(d)*100)/100;
  });

http://nextstrain.org/data/Zika_frequencies.json

there is no ["mutations"], is this the right request url?

For geo, attr, etc.

Maybe:

1. Tree on map
2. ebola like thing
3. pins on leaflet map
4. https://bost.ocks.org/mike/leaflet/

• Colors
• Authors

Only show epitope for flu, for instance

(controls, ie., what things you can click on, what options there are to click on)

It's trivial to precompute radial and unrooted trees. X and Y values from these layouts can be added as node attributes in nextstrain-augur. We can then add buttons to switch layouts in auspice. Animation will be good to have.

Version datasets with augur: /data/43922349852/flu_h2n2_3y_tree.json, etc...

Then keep the version ID 43922349852 in auspice params. This would allow deep-linking back to a previous augur build.

Developing this would work like custom augur builds.