💾 Archived View for thebird.nl › gn-gemtext-threads › issues › authorisation.gmi captured on 2022-07-16 at 15:45:07. Gemini links have been rewritten to link to archived content

-=-=-=-=-=-=-

Clean up Authorisation

Look at Chris' authorisation code:

https://github.com/genenetwork/gn-proxy

Atm genenetwork2's handling of

authorisation is a bit clunky; Since @Bonface is working on the data

data up-loader, and it requires authorisation, this is a good time to

clean up authorisation.

TODO

- [x] Make user able to delete/ add extra value

- [x] Fetch complete list of samples from database and genotype file

instead of only fetching that list from the database. Look at trait

page for reference.

- [ ] Extend idea of csv generation to probeset data. Get data from

the genotype file and work out how to update the values from the

database.

Notes

mRNA datasets ("ProbeSetFreeze" in the DB) contain traits, so when you

assign privileges to an mRNA dataset, it will apply to all traits

within.

But the phenotype resources (for example the dataset name

"BXDPublish"; the stuff in the PublishXRef, etc tables) are just the

traits themselves. The rows of "PublishFreeze" in the DB don't

correspond to individual resources like the rows of ProbeSetFreeze

(neither do databased genotypes - GenoFreeze - for that matter). There

isn't really any "containing" classification for them beyond the

group. BXD phenotype traits can each come from different studies and

thus have different privileges for different users, etc, but there's

nothing in the database really representing that.

This means that if you want to change privileges for phenotype data,

you have to do it trait by trait (or at least to a selection of traits

you manually choose). But for mRNA ("ProbeSet") traits you can just

refer to the entire dataset (and as far as I'm aware this will always

be the case; there's never a need to assign privileges trait-by-trait

for the mRNA traits.

Another way of representing this is that for mRNA datasets you have

the following levels:

Species -> Group -> Dataset (this is a resource) -> Traits

But for phenotypes/genotypes:

Species -> Group -> Traits (each is a resource)

On the UI(quote from Zach):

One interface issue I ran into is how to add/remove groups (since

currently the proxy only cares about groups for retrieving the

privilege masks). I think I set it up so that you can search for

groups by a user's name, e-mail, or parts of the Group Name itself,

though I was uncertain about that because it means that anyone doing

this could see every group/user. For adding users to a group, I

have people direct input e-mails (instead of giving the ability to

search).

Mon 18 Oct 2021

- Marked the function `check_owner_or_admin` as deprecated. It's only

used in: `wqflask/show_trait/show_trait.py`.

- Heavy use of the walrus operator to remove alot of boiler plate.

Tue 19 Oct 2021

- Every resource needs to have the 3 fields set for the group-mask and

default-mask: "data", "metadata", "admin". Otherwise, a hash-ref

error will be thrown:

03 ~~> hash-ref: no value found for key
  key: 'admins

On this regard, the documentation for gn-proxy should be updated

since it's abit misleading.

- An owner, from the proxy views has super-user privileges for

everything

Mon 26 Oct 2021

- Some issues Zach pointed out:

The /trait/update endpoint*, when accessed from >

"/trait/<name>/edit/inbredset-id/<inbredset_id>", currently doesn't

seem to work because kwargs/args are empty. This is presumably due

to it being a POST endpoint and not having the necessary fields

passed through the URL. I can access all the form inputs in the

decorator with request.form, though those form inputs still don't

include the resource_id or inbredset_id/dataset_name with those same

variable names; they're in the edit_phenotype form as inbred-set-id

and dataset-name respectively (this is why I mentioned the variable

names; if I gave the decorator access to them with request.form,

they didn't match the variable names it expected). So I'm not

really sure how best to fix this; I'm guessing that either the >

trait/update endpoint needs to be changed to include the resource ID

> (though this is complicated by the fact that there can be many

traits within a resource_id for ProbeSet traits; it's only phenotype

traits where each individual trait is its own resource), or the

decorator itself needs to be changed to also access request.form.

There's also a separate issue with

"/trait/edit/probeset-name/<dataset_name>" that likely stems from

the fact that you need both a dataset name and trait ID for

"probeset" traits (and in this case dataset_name is actually the

trait ID). So this probably calls for the endpoint itself being

changed in some way.

- Fixed the above, while at the same time, adding the new auth links

to the show trait page. Key point is that I changed the URLs and

moved them to their own module. Particularly for published datasets.