Add support for FLOAT to Python RowCoder

[TheNeuralBit]

Part of #19815

This adds support for encoding the schema type FLOAT to Python's RowCoder. This is added in a new coder, SinglePrecisionFloatCoder, designed to be compatible with Java's FloatCoder.

GitHub Actions Tests Status (on master branch)

See CI.md for more information about GitHub Actions CI.

[TheNeuralBit]

@lostluck it looks like Go might be using the same coder for single-precision floats as for double-precision, is that right? From the Go PreCommit:

13:58:13 --- FAIL: TestStandardCoders (0.03s)
13:58:13 --- FAIL: TestStandardCoders/beam:coder:row:v1#06 (0.00s)
13:58:13 fromyaml_test.go:42: Failed "{beam:coder:row:v1 13:58:13 13:58:13 �f_float�����$8c97b6c5-69e5-4733-907b-26cd8edae612 [] false}": err decoding "\x01\x00\x00\x00\x00\x00": decoding a struct { F_float float32 "beam:\"f_float\"" }
13:58:13 caused by:
13:58:13 error decoding double field
13:58:13 caused by:
13:58:13 EOF
13:58:13 FAIL

[TheNeuralBit]

Yes that does seem to be the case:

beam/sdks/go/pkg/beam/core/graph/coder/row_encoder.go

Lines 211 to 214 in cc0b446

	casereflect.Float32, reflect.Float64:
	returntypeEncoderFieldReflect{encode: func(rv reflect.Value, w io.Writer) error {
	returnEncodeDouble(rv.Float(), w)
	}}, nil

Would you be opposed to me sending a change to encode as single-precision in that case?

[codecov]

Codecov Report

Merging #22626 (bcb3822) into master (7a9bb76) will decrease coverage by 0.00%.
The diff coverage is 80.00%.

@@ Coverage Diff @@## master #22626 +/- ##
==========================================
- Coverage 74.19% 74.19% -0.01% 
==========================================
Files 709 709 Lines 93499 93524 +25 ==========================================
+ Hits 69370 69387 +17 - Misses 22852 22860 +8 
Partials 1277 1277 

Flag	Coverage Δ
python	83.59% <80.00%> (-0.01%)	⬇️

Flags with carried forward coverage won't be shown. Click here to find out more.

Impacted Files	Coverage Δ
...ks/python/apache_beam/coders/coders_test_common.py	98.54% <ø> (ø)
sdks/python/apache_beam/coders/coders.py	88.00% <63.63%> (-0.35%)	⬇️
sdks/python/apache_beam/coders/coder_impl.py	93.91% <85.71%> (-0.06%)	⬇️
sdks/python/apache_beam/coders/row_coder.py	94.64% <100.00%> (+0.14%)	⬆️
sdks/python/apache_beam/coders/slow_stream.py	94.69% <100.00%> (+0.19%)	⬆️
sdks/python/apache_beam/internal/metrics/metric.py	93.00% <0.00%> (-1.00%)	⬇️
...hon/apache_beam/runners/direct/test_stream_impl.py	93.28% <0.00%> (-0.75%)	⬇️
...eam/runners/portability/fn_api_runner/execution.py	92.44% <0.00%> (-0.65%)	⬇️
...ks/python/apache_beam/runners/worker/sdk_worker.py	88.94% <0.00%> (-0.16%)	⬇️
...on/apache_beam/runners/dataflow/dataflow_runner.py	82.87% <0.00%> (-0.14%)	⬇️
... and 3 more

📣 We’re building smart automated test selection to slash your CI/CD build times. Learn more

[github-actions]

Assigning reviewers. If you would like to opt out of this review, comment assign to next reviewer:

R: @AnandInguva for label python.

Available commands:

• stop reviewer notifications - opt out of the automated review tooling
• remind me after tests pass - tag the comment author after tests pass
• waiting on author - shift the attention set back to the author (any comment or push by the author will return the attention set to the reviewers)

The PR bot will only process comments in the main thread (not review comments).

[TheNeuralBit]

Well I tried to modify Go's RowCoder but couldn't manage to get the test working right. Instead I've just skipped this case, and filed #22629 to track.

[TheNeuralBit]

Run Java PreCommit

[TheNeuralBit]

@AnandInguva do you have time to review this?

[AnandInguva]

@TheNeuralBit I am on vacation. I will be back on Aug 22nd.

[TheNeuralBit]

assign to next reviewer

[TheNeuralBit]

R: @tvalentyn

[github-actions]

Stopping reviewer notifications for this pull request: review requested by someone other than the bot, ceding control

[TheNeuralBit]

Oops Valentyn is also out

R: @ryanthompson591

[TheNeuralBit]

Noting there's a race with #22664. Whichever is merged later should remove this exception.

[TheNeuralBit]

@jrmccluskey this test case is still failing in Go. It looks like it may just be a test infra issue, but I'm not able to debug on my own. Could you take a look?

I'm re-opening #22629 to track

[jrmccluskey]

Can you share the output from the test failure?

[TheNeuralBit]

Sorry I should have shared it to begin with but it had scrolled out of my shell history and I was lazy 😬

I dropped the log in #22629 (comment)

[TheNeuralBit]

Run Go PreCommit

[ryanthompson591]

why f_float instead of float or float32?

[TheNeuralBit]

f_float here is the name of a field in the schema. We could call it "float" or "float32" instead, but it's not that important. I just followed the f_ convention used elsewhere.

[ryanthompson591]

can you double check this is not a typo. For example with double this is:
beam:coder:double:v1

I would suspect this should be
beam:coder:float:v1

[TheNeuralBit]

This is a reference to row coder:

beam/model/pipeline/src/main/proto/org/apache/beam/model/pipeline/v1/beam_runner_api.proto

Line 1062 in 3173b50

ROW=13 [(beam_urn) = "beam:coder:row:v1"];

Row Coder is a complex coder that is parameterized by a Beam Schema. It encodes different types of values by deferring to other coders, like beam:coder:double:v1 for DOUBLE