Extend the parser's list of binary operators

changelog.md

@@ -10,6 +10,7 @@ Features
 Bug Fixes
 ---------
 * Suppress warnings when `sqlglotrs` is installed.
+* Improve completions after operators, by recognizing more operators.
 
 
 1.64.0 (2026/03/13)

mycli/packages/completion_engine.py

@@ -17,6 +17,30 @@
     re.IGNORECASE,
 )
 
+# missing because not binary
+#   BETWEEN
+#   CASE
+# missing because parens are used
+#   IN(), and others
+# unary operands might need to have another set
+#   not, !, ~
+# arrow operators only take a literal on the right
+#   and so might need different treatment
+# := might also need a different context
+# sqlparse would call these identifiers, so they are excluded
+#   xor
+# these are hitting the recursion guard, and so not completing after
+# so we might as well leave them out:
+#   is, 'is not', mod
+# sqlparse might also parse "not null" together
+# should also verify how sqlparse parses every space-containing case
+BINARY_OPERANDS = {
+    '&', '>', '>>', '>=', '<', '<>', '!=', '<<', '<=', '<=>', '%',
+    '*', '+', '-', '->', '->>', '/', ':=', '=', '^', 'and', '&&', 'div',
+    'like', 'not like', 'not regexp', 'or', '||', 'regexp', 'rlike',
+    'sounds like', '|',
+}  # fmt: skip
+
 
 def _enum_value_suggestion(text_before_cursor: str, full_text: str) -> dict[str, Any] | None:
     match = _ENUM_VALUE_RE.search(text_before_cursor)
@@ -333,8 +357,6 @@ def suggest_based_on_last_token(
     else:
         token_v = token.value.lower()
 
-    is_operand = lambda x: x and any(x.endswith(op) for op in ["+", "-", "*", "/"])  # noqa: E731
-
     if not token:
         return [{"type": "keyword"}, {"type": "special"}]
 
@@ -512,11 +534,19 @@ def suggest_based_on_last_token(
     elif is_inside_quotes(text_before_cursor, -1) in ['single', 'double']:
         return []
 
-    elif token_v.endswith(",") or is_operand(token_v) or token_v in ["=", "and", "or"]:
+    elif token_v.endswith(",") or token_v in BINARY_OPERANDS:
         original_text = text_before_cursor
         prev_keyword, text_before_cursor = find_prev_keyword(text_before_cursor)
         enum_suggestion = _enum_value_suggestion(original_text, full_text)
-        fallback = suggest_based_on_last_token(prev_keyword, text_before_cursor, None, full_text, identifier) if prev_keyword else []
+
+        # guard against non-progressing parser rewinds, which can otherwise
+        # recurse forever on some operator shapes.
+        if prev_keyword and text_before_cursor.rstrip() != original_text.rstrip():
+            fallback = suggest_based_on_last_token(prev_keyword, text_before_cursor, None, full_text, identifier)
+        else:
+            # perhaps this fallback should include columns
+            fallback = [{"type": "keyword"}]
+
         if enum_suggestion and _is_where_or_having(prev_keyword):
             return [enum_suggestion] + fallback
         return fallback

test/test_completion_engine.py

@@ -126,6 +126,27 @@ def test_operand_inside_function_suggests_cols2():
     assert suggestion == [{"type": "column", "tables": [(None, "tbl", None)]}]
 
 
+def test_operand_inside_function_suggests_cols3():
+    suggestion = suggest_type("SELECT MAX(col1 ||  FROM tbl", "SELECT MAX(col1 || ")
+    assert suggestion == [{"type": "column", "tables": [(None, "tbl", None)]}]
+
+
+def test_operand_inside_function_suggests_cols4():
+    suggestion = suggest_type("SELECT MAX(col1 LIKE  FROM tbl", "SELECT MAX(col1 LIKE ")
+    assert suggestion == [{"type": "column", "tables": [(None, "tbl", None)]}]
+
+
+def test_operand_inside_function_suggests_cols5():
+    suggestion = suggest_type("SELECT MAX(col1 DIV  FROM tbl", "SELECT MAX(col1 DIV ")
+    assert suggestion == [{"type": "column", "tables": [(None, "tbl", None)]}]
+
+
+@pytest.mark.xfail
+def test_arrow_op_inside_function_suggests_nothing():
+    suggestion = suggest_type("SELECT MAX(col1->  FROM tbl", "SELECT MAX(col1->")
+    assert suggestion == []
+
+
 def test_select_suggests_cols_and_funcs():
     suggestions = suggest_type("SELECT ", "SELECT ")
     assert sorted_dicts(suggestions) == sorted_dicts([
@@ -418,6 +439,8 @@ def test_join_alias_dot_suggests_cols2(sql):
     [
         "select a.x, b.y from abc a join bcd b on ",
         "select a.x, b.y from abc a join bcd b on a.id = b.id OR ",
+        "select a.x, b.y from abc a join bcd b on a.id = b.id + ",
+        "select a.x, b.y from abc a join bcd b on a.id = b.id < ",
     ],
 )
 def test_on_suggests_aliases(sql):