[Bug Fix] Broken graphQL query comparisons

[BhaskarChezhiyan]

Overview

graphQL queries with conditions that have an equal or not equal comparison with 0 as operand value are not evaluated properly.

Reproduce

import vaex
import pandas as pd
from decimal import Decimal

df = vaex.from_pandas(pd.Dataframe(data=[[7]], columns=['dummy'] ))

query = """
query($where: BoolExp!, $limit: Int!){
          df(where: $where) { row(limit: $limit){
                               dummy
                               }
                        }
                  }
"""
variables = {'limit': Decimal('10'), 'where': {'dummy': {'_eq': Decimal('0')}}}
output = df.graphql.execute(query, variable_values=variables)

print(output.errors)

Changes

1. Added null check to Compare class _eq and _neq checks.
2. Fixed BooleanCompare class to inherit from Compare class rather than NumberCompare.

[ddelange]

hey 👋

in #2331 we are encountering an issue with graphql syntax when upgrading to graphene v3: graphql-python/graphene-tornado@a63677c

graphene-tornado v2 (currently required by vaex-graphql) pins werkzeug to an incompatible version for python 3.12 (incompatible with vaex-ml / tensorflow 2.16 to be precise), forcing us to upgrade to graphene-tornado==3.0.0b2, which unpins werkzeug but also requires graphene v3.

do you have any idea how to fix this one?

    def test_aggregates(df, schema):
        result = schema.execute("""
        {
            df {
                count
                min {
                    x
                    y
                }
                mean {
                    x
                    y
                }
                max {
                    x
                    y
                }
            }
        }
        """)
>       assert not result.errors
E       assert not [GraphQLError("Aggregation.__init__() got an unexpected keyword argument 'df'", locations=[SourceLocation(line=3, column=9)], path=['df'])]
E        +  where [GraphQLError("Aggregation.__init__() got an unexpected keyword argument 'df'", locations=[SourceLocation(line=3, column=9)], path=['df'])] = ExecutionResult(data={'df': None}, errors=[GraphQLError("Aggregation.__init__() got an unexpected keyword argument 'df'", locations=[SourceLocation(line=3, column=9)], path=['df'])]).errors

edit: we ended up staying with graphene-tornado v2, by manually unpinning werkzeug in the 2.6.1 release (see top level setup.py used in CI). we can't do this in the official vaex-graphql release as PyPI will reject direct references. end-users can do this too by adding to their requirements file or pip install command (with quotes):

graphene-tornado @ https://github.com/ddelange/graphene-tornado/archive/refs/heads/2.6.1.unpin-werkzeug.zip