-
Notifications
You must be signed in to change notification settings - Fork 26
Provenance UI thoughts
We need to support two kinds of actions: get and show phylogeny-related info for every visible edge, and get and show taxonomy-related info for every node that has it.
All this information should be stored in treemachine/neo4j, and accessed via service call by the tree viewer using some as-yet-to-be-determined user interface.
Edge info consists of study info (and thence study info; often they're 1-1) for as many studies as are involved in the edge, up to some limit. There may be many studies (maybe too many to display conveniently), or none (the edge therefore must have come from the taxonomy). For each study we should make available
- The bibliographic information for the study, abbreviated somehow for viewing (first author and year??)
- The DOI or other URL for the article reporting the study, when available (as a hyperlink on the former)
- A treebase or Dryad link when available
- A Phylografter link (always present since all studies come via Phylografter) (Not sure about this one, but better to implement it now and then remove it if we decide not to show it)
As a straw man, consider the following, which might appear in a sidebar as one hovers over an edge (or some other UI - the purpose here is not to propose UI but to show what it needs to do):
11. Magallon et al 2007.
20. Soza et al 2010. (Treebase 10364)
This says that the edge is supported by trees coming from two studies, study 11 and study 20. Click on the 11 or 20 and you go t Phylografter. Click on the author/year and you go to the article reporting the study. Click on the treebase id and you go to treebase.
(Maybe there could be a link (based on plugging the study id into a template) to browse to a single tree - this is a feature Mark H has implemented, I think.)
Question: How much / how should we expose tree-level information? E.g. one would browse to a particular tree, not to a study, which could have many trees. Maybe summary information would be a useful substitute for gruesome detail, e.g. as saying that the edge is present in 3 out of 7 trees reported in the study? Triage: forget about trees, just talk about studies.
Details: Figure out what to do if there are an excessive number of studies associated with an edge. Triage: stop the displayed list at some cutoff (5 or 10) and say "... and 13 others".
Taxon info: From the node, you can get (in addition to the name) an ottol id (for disambiguation), and from there you can get an NCBI or GBIF id, which by prepending a known prefix gets you to a page we can link to.
Other sorts of links might be nice. Look at NCBI for inspiration http://www.ncbi.nlm.nih.gov/Taxonomy/Browser/wwwtax.cgi . Initially just the NCBI and GBIF links are fine.
Linking to GBIF is problematic since the page is behind a very annoying click-through. This argues for some NCBI-like linkouts for the GBIF-originated taxa directly in out UI. I [JAR] don't even know what GBIF has - I've never agreed to the click-through. Postpone dealing with this until more GBIF details are copied into the opentree taxonomy dump.
Currently in argus clicking on a node is a way to navigate through the tree. In onezoom clicking zooms in (or does whatever the currently selected action is). If we continue these uses of single-click then single-click will be unavailable for provenance display, and we'll need a different control for showing the linkout/provenance information and acting on parts of it.
Mouseovers have been suggested. There are many variations. The information (linkouts etc) could be displayed near the node, or over the pane/window. That's pretty intrusive. One idea was to put the information in a fixed pane off to the side or top or bottom that is always there, but whose contents change whenever you mouse over a node or edge. But then getting to that page so you can click on links etc. is a pain.
I'm not a fan of right-click, but that's an option.
Or there could be a little nub that's part of a node, or just next to a node, that takes you to a separate page that has all the linkout info etc. on it.
Maybe taxonomy information could be a hyperlink on the taxonomic name (as in the NCBI taxonomy browser e.g. http://www.ncbi.nlm.nih.gov/Taxonomy/Browser/wwwtax.cgi?id=8399 ), and edge info could come from clicking on an edge.
Design suggestions sought.
UI performance will be better if all edge and node metadata is prefetched (i.e. fetched at the same time that the tree view comes in); without that, mouse motions will initiate network roundtrips, leading to annoying delays. On the other hand the edge information might be voluminous. Decision TBD.