Add feature and feature set labels, for metadata by imjuanleonard · Pull Request #536 · feast-dev/feast

imjuanleonard · 2020-03-16T04:42:15Z

What this PR does / why we need it:
-> Extension for #463

Update (@suwik):

Addressed previous review comments
Added labels on the feature set level
Removed the python SDK changes related to labels implemented inside field.py (30d63fa)
Made a small unit test refactoring on the way.

Which issue(s) this PR fixes:
Fixes #463

Does this PR introduce a user-facing change?:

Feature spec and Feature set spec will both have a new field called labels

feast-ci-bot · 2020-03-16T04:42:30Z

Hi @imjuanleonard. Thanks for your PR.

I'm waiting for a gojek member to verify that this patch is reasonable to test. If it is, they should reply with /ok-to-test on its own line. Until that is done, I will not automatically test new commits in this PR, but the usual testing commands by org members will still work. Regular contributors should join the org to skip this step.

Once the patch is verified, the new status will be reflected by the ok-to-test label.

I understand the commands that are listed here.

Details

Instructions for interacting with me using PR comments are available here. If you have questions or suggestions related to my behavior, please file an issue against the kubernetes/test-infra repository.

imjuanleonard · 2020-03-16T04:45:42Z

Hello @woop I cut the branch from the #499 and latest of #508 and implemented the SDK for the same
When this request gets merge, we can close #499 and #508 altogether

khorshuheng · 2020-03-16T04:50:39Z

/ok-to-test

imjuanleonard · 2020-03-16T04:51:43Z

Implementation Detail
-> Cut Branch from #499
-> Cut Branch from #508
-> Implement the new python SDK for labels

imjuanleonard · 2020-03-16T04:51:56Z

Demo List Features

from feast import Client,Entity, ValueType, FeatureSet, Feature
from pprint import pprint

client = Client(core_url="localhost:6565", serving_url="localhost6566")
client.set_project("test")

client_features = client.list_features()
pprint(vars(client_features['city']))

Output

{'_dtype': <ValueType.STRING: 2>,
 '_labels': {'key_test': 'value_test'},
 '_name': 'city'}

imjuanleonard · 2020-03-16T05:04:21Z

Demo List Feature Set

Requirements
Add this function so it can prety print the feast/feature.py

    def __repr__(self):
        return "<Test name:%s dtype:%s labels:%s>" % (self._name, self.dtype, self.labels)

Create the python function to list feature set and print the features

from feast import Client,Entity, ValueType, FeatureSet, Feature
client = Client(core_url="localhost:6565", serving_url="localhost6566")
client.set_project('juan')
test = client.list_feature_sets(project="juan",name="driver",version="1")
for featureset in test:
    print(featureset.features)

Result:

[<Test name:city dtype:ValueType.STRING labels:{'key_test': 'value_test'}>]

imjuanleonard · 2020-03-16T05:07:48Z

Example Set Labels

from google.protobuf.duration_pb2 import Duration
from feast import Client,Entity, ValueType, FeatureSet, Feature

client = Client(core_url="localhost:6565", serving_url="localhost6566")
client.set_project('juan')

feature1=Feature(name='total_transaction4',dtype=ValueType.INT64, labels=dict())
feature1.set_label("test","tost")
feature1.set_label("tast","tust")
feature1.set_label("bost","bust")

customer_fs = FeatureSet("customer_transactions",
                         entities=[Entity(name='customer_id4', dtype=ValueType.INT64, labels=dict())],
                         features=[feature1],
                         max_age=Duration(seconds=432000))
client.apply(customer_fs)

Result

Feature set updated/created: "customer_transactions:"

DB

feature_set_id          | juan/customer_transactions:5
bool_domain             |
domain                  |
float_domain            |
group_presence          |
image_domain            |
int_domain              |
labels                  | {"tast":"tust","test":"tost","bost":"bust"}
mid_domain              |
name                    | total_transaction4
natural_language_domain |
presence                |
project                 | juan
shape                   |
string_domain           |
struct_domain           |
time_domain             |
time_of_day_domain      |
type                    | INT64
url_domain              |
value_count             |
version                 | 5

imjuanleonard · 2020-03-16T05:59:33Z

/retest

sdk/python/feast/client.py

woop · 2020-03-17T02:55:19Z

sdk/python/feast/field.py

Shouldnt labels have a type?

It is of type dictionary it is handled on the initiation below, this is to make the labels Optional so the function signature doesn't change

You can use Optional[Dict] there I think

If what you mean is typing.Optional

I have tried to use it before, it is not behaving as optional parameter instead according to the documentation it is behaving as optional value

It's not entirely clear what you mean. Can you please refer to this explanation of Optional type hints: https://stackoverflow.com/questions/51710037/how-should-i-use-the-optional-type-hint

Okay, I guess after python 3.5 they are using the Optional aforementioned for type hinting. Have added it

sdk/python/feast/loaders/ingest.py

woop · 2020-03-17T03:00:24Z

@imjuanleonard we absolutely need tests in order to sign this off. Can you please add them in the same PR?

imjuanleonard · 2020-03-17T21:25:51Z

#536 (comment)

The test has been added

sdk/python/feast/feature.py

woop · 2020-03-18T03:15:32Z

sdk/python/feast/field.py

Is there a reason why some code is on the feature and some of it is on the field? Can we label entities? If not, why not?

I am following the current discussion that the use case people need are for the feature currently. So I don't want to assume that people need for the entities as well, Do we need metadata for the entity?

It's not so much about entity vs feature as it is feature vs field. Entities inherit from Fields. If the Field can have a label then an Entity can have a label. So we should either limit this entirely to features, or we should add labels to both Entities and Features, which may happen through the Field class.

Looking at the .proto files on this branch, I can see that labels field is only applied to Features (not Entities or Fields). Not sure how to proceed with this one as I'm not familiar with requirements/use cases of the Python client. Any suggestion?

Actually Fields are the carrier of (feature and entity) values—actual data, not the metadata of registry specifications. I think these changes to field.py should probably be backed out.

From discussion Friday, it sounds like we're in agreement to leave entity labeling out of scope for now. Useful thing to have someday (perhaps when #405 gets focus), but it wasn't initially planned for Feast 0.5. So I think we should proceed with adding map<string, string> labels to FeatureSetSpec as we discussed and providing the API for that.

We might have to defer Python SDK at least in part to the Gojek team, we don't currently use it. It's the primary means of verifying end-to-end functional integration in the project, but maybe as a compromise we can demonstrate the implementation using grpc_cli examples as a test plan, and Python SDK could be finished in a second PR.

@woop Does this sound reasonable?

Btw @suwik have a look through tests/e2e directory in the root of the project for examples of Python SDK, useful for learning the system. They're kicked off for CI by scripts in infra/scripts (used to be in .prow/scripts at the point of our internal fork).

Thanks for suggestion @ches. I'm thinking it would be good to add some logic testing the labels to the e2e tests (and later we could take that to our repo too). I tried to follow the installation instructions in the readme, but after running docker-compose up none of the mentioned ports are open. Any suggestion how to get my environment running for this repo so that I'm able to run the e2e tests?

EDIT: ok, seems like I had COMPOSE_FILE set to our internal fork file and that's why it wasn't starting all the services that I was expecting to.

I have removed the python SDK changes related to labels (implemented inside field.py). I did that in a separate commit so it should be easy to rollback in case we actually need them.

sdk/python/feast/loaders/ingest.py

ches · 2020-03-22T10:37:32Z

I was a bit misled by the PR title and trying to follow the history from #499—this PR is not only Python SDK, it's complete implementation of feature labels.

Conflicts need to be fixed now (mostly generated files that have been removed from Git)—if in the process you could rebase and squash some of the false-start commits like list_features functionality that was removed, and typos, formatting, and comment updates, that would be fantastic. Not absolutely essential for you to do it, but being a significant PR, I'd love to be able to review it commit-by-commit.

ches · 2020-04-29T05:57:14Z

core/src/main/java/feast/core/model/Field.java


+  // Labels that this field belongs to
+  @Column(name = "labels", columnDefinition = "text")
+  private String labels;


Hmm so I guess we're going to want to move this to the Feature entity class in #655… this was mentioned before but I guess I wasn't understanding the implications until I saw #655 reduced from #612.

Creates a bit of a conundrum for us since I thought the labels feature could be a straightforward backport (internal, assuming Feast isn't going to backport features). That's not going to be the case if this has to wait on #655 and substantial SQL schema migrations are going to be required for it.

woop · 2020-04-29T13:19:14Z

/test test-end-to-end-batch

woop · 2020-04-29T13:20:17Z

Looks good to me. Will you let me know when its ready to be merged?

woop · 2020-04-29T13:23:29Z

Thanks for being so patient here @ches. I guess the conversation around Field, Entity, and Feature has temporarily been answered, although I personally don't think we have reached our destination yet. I do feel like the door is left more open with our current approach.

ches · 2020-04-29T20:08:42Z

core/src/test/java/feast/core/service/SpecServiceTest.java

+    Map<String, String> featureSetLabels =
+        new HashMap<>() {
+          {
+            put("description", "My precious feature set");


Good spot! ;)

…ard/feast into python-sdk-for-labels

zhilingc · 2020-04-30T06:46:58Z

tests/e2e/grpc-based-registration.py

+LABEL_KEY = "my"
+LABEL_VALUE = "label"
+
+


The tests look good, but possibly add a TODO to migrate this over to the python sdk once that has been updated?

suwik · 2020-04-30T07:57:07Z

/test test-end-to-end

zhilingc · 2020-04-30T08:03:08Z

tests/e2e/basic-ingest-redis-serving.py

+        return get_feature_set_response.feature_set
+
+    @pytest.mark.timeout(45)
+    @pytest.mark.run(order=9)


I think @woop uses these to group similar tests together (10- - so maybe your tests can start from 50?

Perhaps this stuff can be tweaked when changing the tests to use SDK in #663

suwik · 2020-04-30T08:31:35Z

/test test-end-to-end

feast-ci-bot · 2020-04-30T08:56:14Z

[APPROVALNOTIFIER] This PR is APPROVED

This pull-request has been approved by: imjuanleonard, woop

The full list of commands accepted by this bot can be found here.

The pull request process is described here

Details

Needs approval from an approver in each of these files:

~~OWNERS~~ [woop]

Approvers can indicate their approval by writing /approve in a comment
Approvers can cancel approval by writing /approve cancel in a comment

woop · 2020-04-30T08:56:42Z

/lgtm

Backports feast-dev#536 to v0.3

* Add feature and feature set labels Backports #536 to v0.3 * Fix feature labels test for round trip It appears that this test was only exercising the test setup code, not covering that labels work when applied. This change should be forward-ported to master. Co-authored-by: Suwinski, Krzysztof (Agoda) <Krzysztof.Suwinski@agoda.com>

imjuanleonard requested review from davidheryanto, khorshuheng, pradithya, woop and zhilingc as code owners March 16, 2020 04:42

feast-ci-bot added the needs-ok-to-test label Mar 16, 2020

feast-ci-bot added the size/XXL label Mar 16, 2020

imjuanleonard force-pushed the python-sdk-for-labels branch from cced904 to 359fd3d Compare March 16, 2020 04:49

feast-ci-bot added ok-to-test and removed needs-ok-to-test labels Mar 16, 2020

woop reviewed Mar 17, 2020

View reviewed changes

sdk/python/feast/client.py Outdated Show resolved Hide resolved

woop reviewed Mar 17, 2020

View reviewed changes

sdk/python/feast/loaders/ingest.py Outdated Show resolved Hide resolved

imjuanleonard force-pushed the python-sdk-for-labels branch from fa1f75d to 33be193 Compare March 17, 2020 20:20

imjuanleonard mentioned this pull request Mar 17, 2020

Labels for feature #499

Closed

woop reviewed Mar 18, 2020

View reviewed changes

sdk/python/feast/feature.py Outdated Show resolved Hide resolved

woop reviewed Mar 18, 2020

View reviewed changes

sdk/python/feast/loaders/ingest.py Outdated Show resolved Hide resolved

imjuanleonard force-pushed the python-sdk-for-labels branch from 84486a5 to 505d8f7 Compare March 22, 2020 22:06

ches reviewed Apr 29, 2020

View reviewed changes

woop mentioned this pull request Apr 29, 2020

Update Python SDK to support labels #663

Closed

ches reviewed Apr 29, 2020

View reviewed changes

Krzysztof Suwinski added 2 commits April 30, 2020 13:06

added e2e tests for feature and feature set labels

18edc12

Merge branch 'python-sdk-for-labels' of https://github.com/imjuanleon…

5e10c94

…ard/feast into python-sdk-for-labels

feast-ci-bot added size/XL and removed size/L labels Apr 30, 2020

zhilingc reviewed Apr 30, 2020

View reviewed changes

moved e2e tests for feature and feature set labels

02197f8

zhilingc reviewed Apr 30, 2020

View reviewed changes

ches changed the title ~~Add labels as metadata to features~~ Add feature and feature set labels, for metadata Apr 30, 2020

changed tests ordering

8d6d4f4

woop approved these changes Apr 30, 2020

View reviewed changes

feast-ci-bot added the approved label Apr 30, 2020

feast-ci-bot assigned woop Apr 30, 2020

feast-ci-bot added the lgtm label Apr 30, 2020

feast-ci-bot merged commit ea4e2b0 into feast-dev:master Apr 30, 2020

ches mentioned this pull request May 1, 2020

Move TFDV stats to higher-numbered protobuf fields #669

Merged

Joostrothweiler mentioned this pull request May 17, 2020

Add Python SDK support for labels on feature sets #707

Merged

ches mentioned this pull request May 23, 2020

v0.3 backport: Add feature and feature set labels #737

Merged

ches pushed a commit to agoda-com/feast that referenced this pull request May 23, 2020

Add feature and feature set labels

2a9d672

Backports feast-dev#536 to v0.3

		LABEL_KEY = "my"
		LABEL_VALUE = "label"

Conversation

imjuanleonard commented Mar 16, 2020 • edited by suwik Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

feast-ci-bot commented Mar 16, 2020

Uh oh!

imjuanleonard commented Mar 16, 2020

Uh oh!

khorshuheng commented Mar 16, 2020

Uh oh!

imjuanleonard commented Mar 16, 2020

Uh oh!

imjuanleonard commented Mar 16, 2020

Uh oh!

imjuanleonard commented Mar 16, 2020

Uh oh!

imjuanleonard commented Mar 16, 2020

Uh oh!

imjuanleonard commented Mar 16, 2020

Uh oh!

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Uh oh!

woop commented Mar 17, 2020

Uh oh!

imjuanleonard commented Mar 17, 2020

Uh oh!

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

woop Mar 28, 2020 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Choose a reason for hiding this comment

Uh oh!

suwik Apr 26, 2020 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

suwik Apr 27, 2020 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Uh oh!

ches commented Mar 22, 2020

Uh oh!

ches Apr 29, 2020 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Choose a reason for hiding this comment

Uh oh!

woop commented Apr 29, 2020

Uh oh!

woop commented Apr 29, 2020

Uh oh!

woop commented Apr 29, 2020

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

imjuanleonard commented Mar 16, 2020 •

edited by suwik

Loading

woop Mar 28, 2020 •

edited

Loading

suwik Apr 26, 2020 •

edited

Loading

suwik Apr 27, 2020 •

edited

Loading

ches Apr 29, 2020 •

edited

Loading