fix(expressions): keep tabs inside string literals

pyparsing expands tabs in the input before matching, so a tab inside a string
literal was rewritten into spaces, and the number of spaces depended on where
the literal sat in the expression: parse("a = 'x\ty'") and parse("ab = 'x\ty'")
returned different values for the same literal. A filter on a value containing
a tab silently matched nothing.

Author: Rodrigo-Palma

pyiceberg/expressions/parser.py

@@ -298,6 +298,10 @@ def handle_or(result: ParseResults) -> Or:
     ],
 ).set_name("expr")
 
+# pyparsing expands tabs in the input by default, which would rewrite a tab inside a string literal
+# into spaces, and the number of spaces would depend on where the literal sits in the expression.
+boolean_expression.parse_with_tabs()
+
 
 def parse(expr: str) -> BooleanExpression:
     """Parse a boolean expression."""

tests/expressions/test_parser.py

@@ -294,3 +294,14 @@ def test_boolean_as_operand(expression: str, expected: BooleanExpression) -> Non
 def test_boolean_as_literal_is_unchanged() -> None:
     assert parser.parse("foo = true") == EqualTo(Reference("foo"), literal(True))
     assert parser.parse("foo in (true, false)") == In(Reference("foo"), {literal(True), literal(False)})
+
+
+def test_literal_with_tab_is_preserved() -> None:
+    assert EqualTo("foo", "a\tb") == parser.parse("foo = 'a\tb'")
+
+
+def test_literal_with_tab_does_not_depend_on_its_position() -> None:
+    one_char_column = parser.parse("a = 'x\ty'")
+    two_char_column = parser.parse("ab = 'x\ty'")
+
+    assert one_char_column.literal.value == two_char_column.literal.value