Add Docker Hub categories

[yosifkit]

Docker Hub allows owners to set categories for their repos (https://docs.docker.com/docker-hub/repos/categories/), so let's allow our image maintainers in DOI to do it. Categories are not yet visible to users.

To our image maintainers: these initial categories are based on some rough google searching of the contained software and content.md, we apologize for any that are inaccurate. Docker Hub currently allows setting up to 3, so if you'd like to add more, just add them to the list in [repo]/metadata.json.

[yosifkit]

For now, get-categories.sh will be manually run since the available categories shouldn't change that often.

[LaurentGoderre]

This should also have "web-servers"

[LaurentGoderre]

Maybe "developer-tools" like for mongo express

[tianon]

I only reviewed the implementation for now (I have not re-reviewed the categories yet like I plan to), and have a few thoughts/suggestions.

I'm definitely really concerned about having canonical human-maintained JSON as an input (since JSON is not a terribly human or diff friendly format), but it's probably fine for now and I think there are future benefits to having this as JSON. 👍

[whalelines]

Still working through metadata.sh, but submitting this so it does not get lost

[whalelines]

Should we provide example usages?

$ ./metadata.sh -f -c REPO_NAME

[yosifkit]

-f was changed to -w and now I dropped the -c and -d flags (commit incoming with script and README adjustments).

[whalelines]

A reminder that the name property is not used

[yosifkit]

Yeah, it is required to be non-empty, but is unused as far as I can tell.

[whalelines]

There is a -f option, right?

[whalelines]

Suggested change

	while true; do
	while :; do

[whalelines]

I guess there is no -f option. Perhaps the README needs updated?

[tianon]

This should be part of the if statement just above right? (so we're not checking canonicalMetadataFile against itself?)

Could probably then collapse the two jq invocations into one, but getting decent output out of that might be annoying? Maybe have jq generate the error message string contents instead?

[yosifkit]

Before and after most recent commit unifying the jq expression for categories (with some local broken files):

$ time ./metadata.sh */ .
error: test/metadata.json does not exist or is empty
--- unit/metadata.json  2024-05-16 15:50:54.308055944 -0700
+++ /dev/fd/63  2024-05-16 15:50:57.018164177 -0700
@@ -1,9 +1,6 @@
 {
   "hub": {
     "categories": [
-      "web-servers",
-      "web-servers",
-      "web-servers",
       "servers",
       "web-servers"
     ]
error: unit/metadata.json: too many categories: 5 (max 3)
error: unit/metadata.json: unkown categories: ["servers"]

real    0m6.432s
user    0m6.110s
sys     0m0.258s

$ time ./metadata.sh */ .
error: test/metadata.json does not exist or is empty
--- unit/metadata.json  2024-05-16 15:50:54.308055944 -0700
+++ /dev/fd/63  2024-05-16 15:51:15.227048108 -0700
@@ -1,9 +1,6 @@
 {
   "hub": {
     "categories": [
-      "web-servers",
-      "web-servers",
-      "web-servers",
       "servers",
       "web-servers"
     ]
error: unit/metadata.json: too many categories: 5 (max 3)
error: unit/metadata.json: unknown categories ["servers"]

real    0m4.327s
user    0m4.145s
sys     0m0.162s

[tianon]

A few category adjustments (some wrong, some with a better match).

FWIW, I used for r in */; do printf '%s\n%s\n%s\n\n' "${r%/}" "$(jq -r '.hub.categories | join(" ")' "$r/metadata.json")" "$(< "$r/README-short.txt")"; done in a local checkout to review these with slightly more context. 😄

[tianon]

Suggested change

	"security"
	"api-management"

API Management: Tools for creating, publishing, analyzing, and securing APIs.

[tianon]

Maybe we should link to https://docs.docker.com/docker-hub/repos/categories/ in here somewhere? (esp. since it has the descriptions for all the categories)

[tianon]

Suggested change

	"operating-systems"
	"web-servers"

[tianon]

Suggested change

	"security"
	"api-management"

SATOSA translates between authentication protocols such as SAML2, OpenID Connect, and OAuth2.

(you could argue for both, but I think API management is a more accurate starting point)

[tianon]

Suggested change

	"machine-learning-and-ai"
	"data-science"

[tianon]

Suggested change

	"categories": []
	"categories": [
	"security"
	]

[tianon]

Suggested change

	"message-queues"
	"databases-and-storage"

(it takes a bit to unpack the tech lingo on many official zookeeper descriptions, but it's a distributed key-value store, not a message queue 👀)

[whalelines]

Looks good. A couple minor things

[whalelines]

Suggested change

	ie: $self debian
	eg: $self debian

[whalelines]

Suggested change

	usage: $self [--write] repo [...]
	usage: $self [--write] REPO[...]

[whalelines]

LGTM