Corrupt proto files

[andrewthad]

Here is an example of an attempt to filter on a u64 column in an Apache Arrow file:

version:
  majorNumber: 0
  minorNumber: 57
  patchNumber: 1
extensionUris:
- extensionUriAnchor: 1
  uri: https://raw.githubusercontent.com/apache/arrow/refs/heads/main/format/substrait/extension_types.yaml
- extensionUriAnchor: 2
  uri: https://raw.githubusercontent.com/substrait-io/substrait/refs/heads/main/extensions/functions_comparison.yaml
extensions:
- extensionType:
    typeAnchor: 100
    name: 'u32'
    extensionUriReference: 1
- extensionType:
    typeAnchor: 101
    name: 'u64'
    extensionUriReference: 1
- extensionFunction:
    functionAnchor: 200
    name: 'gt'
    extensionUriReference: 2
relations:
- root:
    names:
      - "source_ip"
      - "source_bytes"
    input:
      read:
        namedTable:
          names:
          - traffic
        baseSchema:
          names:
          - source_ip
          - source_bytes
          struct:
            nullability: NULLABILITY_REQUIRED
            types:
            - userDefined:
                typeReference: 100
                nullability: NULLABILITY_REQUIRED
            - userDefined:
                typeReference: 101
                nullability: NULLABILITY_REQUIRED

When I validate with ``, I get a bunch of warnings like this:

Warning (code 0001):
  at plan.extension_uris[1].uri.data.scalar_functions[22]:
  not yet implemented: the following child nodes were not recognized by the validator: description, impls, name (code 0001)

Which seems fine. But when I try to go one step further and convert this yaml to a protobuf, it gets weird. Here's the command I'm using:

../bin/substrait-validator --uri-depth 2 --out-type proto --out-file simpletraffic.proto simpletraffic.yaml

Same warnings, but the proto file that substrait-validator creates is corrupt. It won't round trip with ../bin/substrait-validator --uri-depth 2 simpletraffic.proto, failing with:

Error (code 1001):
  at plan:
  failed to decode Protobuf message: SimpleExtensionUri.extension_uri_anchor: Plan.extension_uris: invalid wire type: LengthDelimited (expected Varint) (code 1001)

Error (code 1002):
  at plan:
  missing required protobuf field: version (code 1002)

Fatal error:
  plan is invalid

Checking the proto file with xxd, I can see that it looks like error and warning messages are interleaved with the binary data:

000055b0: 796e 6f74 2079 6574 2069 6d70 6c65 6d65  ynot yet impleme
000055c0: 6e74 6564 3a20 7468 6520 666f 6c6c 6f77  nted: the follow
000055d0: 696e 6720 6368 696c 6420 6e6f 6465 7320  ing child nodes 
000055e0: 7765 7265 206e 6f74 2072 6563 6f67 6e69  were not recogni
000055f0: 7a65 6420 6279 2074 6865 2076 616c 6964  zed by the valid
00005600: 6174 6f72 3a20 6465 7363 7269 7074 696f  ator: descriptio
00005610: 6e2c 2069 6d70 6c73 2c20 6e61 6d65 2028  n, impls, name (
00005620: 636f 6465 2030 3030 3129 2a4b 0a04 706c  code 0001)*K..pl
00005630: 616e 1214 1212 0a0e 6578 7465 6e73 696f  an......extensio
00005640: 6e5f 7572 6973 1001 1207 0a05 0a03 7572  n_uris........ur

That's not what I would expect.