hubmapconsortium-graveyard / directory-schema Goto Github PK

... perhaps drop relative, and make them absolute links to github?

Specifying directory structures has come up in a couple different contexts (default for zarr, and for datasets, of whatever type, etc.).

I propose that the formal definition of these structures be JSON schema targeting the output of tree -J . , which looks like:

[
  {"type":"directory","name":".","contents":[
    {"type":"file","name":"CHANGELOG.md"},
    {"type":"file","name":"LICENSE"},
    {"type":"file","name":"README.md"},

In the schema, items and contains will be be used on the lists, and regexes will be used on the filenames.

(It will take a few lines of python to emulate tree -J : I don’t see anything that does it out of the box.)

Trevor:

I think this makes sense for our zarr purposes. Zarr is restrictive enough that files are limited to .zgroup, .zarray, .zattrs, and the binary files, but arrays and groups are each named directories and having something to parse a zarr store and say “Yes this is a valid pyramid” would be great.