Query.Farm SQL Manipulation

A Python library for intelligent SQL predicate manipulation using SQLGlot

Column Filtering with Complex Expressions

import sqlglot
from query_farm_sql_manipulation import transforms

sql = '''
SELECT * FROM users
WHERE age > 18
  AND (status = 'active' OR role = 'admin')
  AND department IN ('engineering', 'sales')
'''

# Parse the statement first
statement = sqlglot.parse_one(sql, dialect="duckdb")

# Only keep predicates involving 'age' and 'role'
allowed_columns = {'age', 'role'}

result = transforms.filter_column_references(
    statement=statement,
    selector=lambda col: col.name in allowed_columns,
)

# Result: SELECT * FROM users WHERE age > 18 AND role = 'admin'
print(result.sql())

Features

• Predicate Removal: Safely remove specific predicates from complex SQL WHERE clauses while preserving logical structure
• Column Filtering: Filter SQL statements to only include predicates referencing allowed columns
• Intelligent Logic Handling: Properly handles AND/OR logic, nested expressions, CASE statements, and parentheses
• SQLGlot Integration: Built on top of SQLGlot for robust SQL parsing and manipulation
• Multiple Dialect Support: Works with various SQL dialects (default: DuckDB)

Installation

pip install query-farm-sql-manipulation

Requirements

• Python >= 3.12
• SQLGlot >= 26.33.0

Quick Start

Basic Predicate Removal

import sqlglot
from query_farm_sql_manipulation import transforms

# Parse a SQL statement
sql = 'SELECT * FROM data WHERE x = 1 AND y = 2'
statement = sqlglot.parse_one(sql, dialect="duckdb")

# Find the predicate you want to remove
predicates = list(statement.find_all(sqlglot.expressions.Predicate))
target_predicate = predicates[0]  # x = 1

# Remove the predicate
transforms.remove_expression_part(target_predicate)

# Result: SELECT * FROM data WHERE y = 2
print(statement.sql())

Column-Name Based Filtering

import sqlglot
from query_farm_sql_manipulation import transforms

# Parse SQL statement first
sql = 'SELECT * FROM data WHERE color = "red" AND size > 10 AND type = "car"'
statement = sqlglot.parse_one(sql, dialect="duckdb")

# Filter to only include predicates with allowed columns
allowed_columns = {"color", "type"}

filtered = transforms.filter_column_references(
    statement=statement,
    selector=lambda col: col.name in allowed_columns,
)

# Result: SELECT * FROM data WHERE color = "red" AND type = "car"
print(filtered.sql())

API Reference

remove_expression_part(child: sqlglot.Expression) -> None

Removes the specified SQLGlot expression from its parent, respecting logical structure.

Parameters:

• child: The SQLGlot expression to remove

Raises:

• ValueError: If the expression cannot be safely removed

Supported Parent Types:

• AND/OR expressions: Replaces parent with the remaining operand
• WHERE clauses: Removes the entire WHERE clause if it becomes empty
• Parentheses: Recursively removes the parent
• NOT expressions: Removes the entire NOT expression
• CASE statements: Removes conditional branches

filter_column_references(*, statement: sqlglot.Expression, selector: Callable[[sqlglot.expressions.Column], bool]) -> sqlglot.Expression

Filters a SQL statement to remove predicates containing columns that don't match the selector criteria.

Parameters:

• statement: The SQLGlot expression to filter
• selector: A callable that takes a Column and returns True if it should be preserved, False if it should be removed

Returns:

• Filtered SQLGlot expression with non-matching columns removed

Raises:

• ValueError: If a column can't be cleanly removed due to interactions with allowed columns

where_clause_contents(statement: sqlglot.expressions.Expression) -> sqlglot.expressions.Expression | None

Extracts the contents of the WHERE clause from a SQLGlot expression.

Parameters:

• statement: The SQLGlot expression to extract from

Returns:

• The contents of the WHERE clause, or None if no WHERE clause exists

filter_predicates_with_right_side_column_references(statement: sqlglot.expressions.Expression) -> sqlglot.Expression

Filters out predicates that have column references on the right side of comparisons.

Parameters:

• statement: The SQLGlot expression to filter

Returns:

• Filtered SQLGlot expression with right-side column reference predicates removed

Examples

Complex Logic Handling

The library intelligently handles complex logical expressions:

# Original: (x = 1 AND y = 2) OR z = 3
# Remove y = 2: x = 1 OR z = 3

# Original: NOT (x = 1 AND y = 2)
# Remove x = 1: NOT y = 2 (which becomes y <> 2)

# Original: CASE WHEN x = 1 THEN 'yes' WHEN x = 2 THEN 'maybe' ELSE 'no' END
# Remove x = 1: CASE WHEN x = 2 THEN 'maybe' ELSE 'no' END

Column Filtering with Complex Expressions

sql = '''
SELECT * FROM users
WHERE age > 18
  AND (status = 'active' OR role = 'admin')
  AND department IN ('engineering', 'sales')
'''

# Only keep predicates involving 'age' and 'role'
allowed_columns = {'age', 'role'}

result = transforms.filter_column_references_statement(
    sql=sql,
    selector=lambda col: col.name in allowed_columns,
)

# Result: SELECT * FROM users WHERE age > 18 AND role = 'admin'

Error Handling

The library will raise ValueError when predicates cannot be safely removed:

import sqlglot
from query_farm_sql_manipulation import transforms

# This will raise ValueError because x = 1 is part of a larger expression
sql = "SELECT * FROM data WHERE result = (x = 1)"
statement = sqlglot.parse_one(sql, dialect="duckdb")

# Cannot remove x = 1 because it's used as a value, not a predicate
# This would raise ValueError if attempted

Supported SQL Constructs

• Logical Operators: AND, OR, NOT
• Comparison Operators: =, <>, <, >, <=, >=, LIKE, IN, IS NULL, etc.
• Complex Expressions: CASE statements, subqueries, function calls
• Nested Logic: Parentheses and nested boolean expressions
• Multiple Dialects: DuckDB, PostgreSQL, MySQL, SQLite, and more via SQLGlot

Testing

Run the test suite:

pytest src/query_farm_sql_manipulation/test_transforms.py

The test suite includes comprehensive examples of:

• Basic predicate removal scenarios
• Complex logical expression handling
• Error cases and edge conditions
• Column filtering with various SQL constructs

Contributing

This project uses:

• Rye for dependency management
• pytest for testing
• mypy for type checking
• ruff for linting

Author

This Python module was created by Query.Farm.

License

MIT Licensed.