-
Notifications
You must be signed in to change notification settings - Fork 2.7k
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
SOLR-15118: Convert /v2/collections APIs to annotations #2281
SOLR-15118: Convert /v2/collections APIs to annotations #2281
Conversation
Solr supports two different ways to write v2 APIs: a JSON spec based approach, and one based on annotated POJOs. The POJO method is now preferred. This commit switches the /v2/collections APIs over to the annotation-based approach.
return; | ||
} | ||
|
||
toFlatten.forEach((k, v) -> destination.put(additionalPrefix + k, v)); |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Maybe we should make this recursive to support potentially nested v2 payloads?
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
I actually tried this initially but found the result to be a lot like the code in CollectionsApiMapping.Meta
, which I personally find tough to read with all the map-literals for renaming properties and prefixing properties. I guessed it was more readable to just call this non-recursive version twice. That wouldn't work as well on APIs with deeper nesting, but seemed feasible here.
I'm open to going back that direction though if you prefer it. Did you have a particular signature in mind?
* | ||
* Analogous to the request parameters for v1 /admin/collections?action=BACKUP API. | ||
*/ | ||
public class BackupCollectionBody implements ReflectMapWriter { |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
What do you think about renaming all of these to use *Payload
instead of *Body
?
@JsonProperty | ||
public Boolean perReplicaState; | ||
|
||
public static class RouterInfo implements ReflectMapWriter { |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Router spec may in fact contain arbitrary properties in addition to these two - see DocRouter.getRouterSpec
. Currently only name
and field
are used in the routers that Solr provides but other users may use some other properties in their implementations.
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Good catch. There's a Jackson annotation made for this case (JsonAnyGetter
) - named properties will be mapped as usual but "extra" ones are put in the catchall Map. But it looks like we're not actually using Jackson here to avoid SolrJ needing to depend on it. The JsonProperty annotation is a copy in our own code to avoid the dep.
I could create clone JsonAnyGetter the way we've done with JsonProperty, but idt it's worth it. I'll just remove this RouterInfo class and handle all the properties as a Map for the translation.
Might be worth considering whether a Jackson-dep is worth it after all going forward though.
Thanks for picking this up @gerlowskija Please let me know if you need any help |
public static final String V2_RESTORE_CMD = "restore-collection"; | ||
public static final String V2_CREATE_ALIAS_CMD = "create-alias"; | ||
public static final String V2_SET_ALIAS_PROP_CMD = "set-alias-property"; | ||
public static final String V2_DELETE_ALIAS_CMD = "delete-alias"; |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Should we also add other collection commands here, such as: COLLECTIONPROP, CREATE / DELETESHARD, ADD / MOVE / DELETEREPLICA, MODIFYCOLLECTION, REINDEXCOLLECTION, RENAME, SPLITSHARD ?
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Yes, please
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
They should definitely exist somewhere, but they're not /v2/collections
APIs, so I didn't do that in this PR. If you guys really want them, I'm not against adding those constants in this PR.
But even then they should probably live somewhere else, shouldn't they? (since this class is for /v2/collections
APIs, and those commands are for other APIs? AFAIK this file is only for those APIs that specifically live under /v2/collections
, right?)
Hey, thanks for offering @noblepaul . I do have one thing I was puzzling over: How is SolrJ's There are some annotation-based v2 APIs whose SolrJ request classes implement V2RequestSupport (e.g. CollectionAdminRequest.ClusterProp), so there should be logic somewhere to handle that I imagine? Unless getV2Request will blow up now if called on certain CollectionAdminRequest objects? More broadly I wonder whether there's any value in getV2Request going forward. It never really "took off" to cover APIs other than CollectionAdmin, and there might be better ways to let SolrJ users dip their toes into the v2 waters (e.g. dedicated request classes that better align with the v2 paths/concepts? idk). If we don't have an approach for V2RequestSupport + annotation-based APIs worked out, and there's not strong feeling about carrying V2RequestSupport forward, I'd lean towards making this PR a master-only change, and deprecating getV2Request in 8.9 so it can be removed by this PR on master. |
The V2RequestSupport interface is incompatible with annotated-POJO-based v2 APIs. (At least, without a good bit of code duplication.) Since some of CollectionAdmin commands covered by CollectionAdminRequest now have annotated-POJO-based v2 equivalents, CAR is better off not implementing V2RequestSupport.
Going to merge this with the strategy for dealing with V2RequestSupport that I mentioned in my last comment. If you eventually get a chance to reply to that comment above @noblepaul , lmk if you prefer some other approach and I can pick it back up then. |
Solr supports two different ways to write v2 APIs: a JSON spec based approach, and one based on annotated POJOs. The POJO method is now preferred. This commit switches the /v2/collections APIs over to the annotation-based approach. Since V2RequestSupport only works with jsonspec-based APIs, this commit also changes CollectionAdminRequest to no longer implement that interface.
Solr supports two different ways to write v2 APIs: a JSON spec based approach, and one based on annotated POJOs. The POJO method is now preferred. This commit switches the /v2/collections APIs over to the annotation-based approach. Since V2RequestSupport only works with jsonspec-based APIs, this commit also changes CollectionAdminRequest to no longer implement that interface.
Description
Solr supports two different ways to write v2 APIs: a JSON spec based
approach, and one based on annotated POJOs. The POJO method is now
preferred.
This commit switches the /v2/collections APIs over to the
annotation-based approach.
Solution
This commit introduces a relatively rote migration of the JSON spec over to annotated POJOs. There's little of note here, except for cases where the v2 JSON spec was missing support for params supported by the v1 APIs. Here I had to make a few decisions as to whether property names should be retained, etc. A good example of this is the v2 backup API, which had several params not covered by the existing v2 spec.
Tests
Adds a new test:
V2CollectionsAPIMappingTest.java
to test the v2->v1 conversion.Checklist
Please review the following and check all that apply:
master
branch../gradlew check
.