str_find

Author: chelsea-lin

bigframes/core/compile/sqlglot/expressions/unary_compiler.py

@@ -182,6 +182,29 @@ def _(op: ops.StrContainsRegexOp, expr: TypedExpr) -> sge.Expression:
     return sge.RegexpLike(this=expr.expr, expression=sge.convert(op.pat))
 
 
+@UNARY_OP_REGISTRATION.register(ops.StrFindOp)
+def _(op: ops.StrFindOp, expr: TypedExpr) -> sge.Expression:
+    if op.end is not None:
+        # BigQuery's INSTR doesn't support `end`, so we need to use SUBSTR.
+        # Also, INSTR is 1-based, so we need to adjust the start position.
+        return sge.func(
+            "INSTR",
+            sge.Substring(
+                this=expr.expr,
+                start=sge.convert(op.start + 1) if op.start is not None else None,
+                length=sge.convert(op.end - (op.start or 0)),
+            ),
+            sge.convert(op.substr),
+        ) - sge.convert(1)
+    else:
+        return sge.func(
+            "INSTR",
+            expr.expr,
+            sge.convert(op.substr),
+            sge.convert(op.start + 1) if op.start is not None else None,
+        ) - sge.convert(1)
+
+
 @UNARY_OP_REGISTRATION.register(ops.StrContainsOp)
 def _(op: ops.StrContainsOp, expr: TypedExpr) -> sge.Expression:
     return sge.Like(this=expr.expr, expression=sge.convert(f"%{op.pat}%"))

tests/unit/core/compile/sqlglot/expressions/snapshots/test_unary_compiler/test_str_find/out.sql

@@ -0,0 +1,13 @@
+WITH `bfcte_0` AS (
+  SELECT
+    `string_col` AS `bfcol_0`
+  FROM `bigframes-dev`.`sqlglot_test`.`scalar_types`
+), `bfcte_1` AS (
+  SELECT
+    *,
+    INSTR(`bfcol_0`, 'e', 1) - 1 AS `bfcol_1`
+  FROM `bfcte_0`
+)
+SELECT
+  `bfcol_1` AS `string_col`
+FROM `bfcte_1`

tests/unit/core/compile/sqlglot/expressions/snapshots/test_unary_compiler/test_str_find/out_with_end.sql

@@ -0,0 +1,13 @@
+WITH `bfcte_0` AS (
+  SELECT
+    `string_col` AS `bfcol_0`
+  FROM `bigframes-dev`.`sqlglot_test`.`scalar_types`
+), `bfcte_1` AS (
+  SELECT
+    *,
+    INSTR(SUBSTRING(`bfcol_0`, 1, 5), 'e') - 1 AS `bfcol_1`
+  FROM `bfcte_0`
+)
+SELECT
+  `bfcol_1` AS `string_col`
+FROM `bfcte_1`

tests/unit/core/compile/sqlglot/expressions/snapshots/test_unary_compiler/test_str_find/out_with_start.sql

@@ -0,0 +1,13 @@
+WITH `bfcte_0` AS (
+  SELECT
+    `string_col` AS `bfcol_0`
+  FROM `bigframes-dev`.`sqlglot_test`.`scalar_types`
+), `bfcte_1` AS (
+  SELECT
+    *,
+    INSTR(`bfcol_0`, 'e', 3) - 1 AS `bfcol_1`
+  FROM `bfcte_0`
+)
+SELECT
+  `bfcol_1` AS `string_col`
+FROM `bfcte_1`

tests/unit/core/compile/sqlglot/expressions/snapshots/test_unary_compiler/test_str_find/out_with_start_and_end.sql

@@ -0,0 +1,13 @@
+WITH `bfcte_0` AS (
+  SELECT
+    `string_col` AS `bfcol_0`
+  FROM `bigframes-dev`.`sqlglot_test`.`scalar_types`
+), `bfcte_1` AS (
+  SELECT
+    *,
+    INSTR(SUBSTRING(`bfcol_0`, 3, 3), 'e') - 1 AS `bfcol_1`
+  FROM `bfcte_0`
+)
+SELECT
+  `bfcol_1` AS `string_col`
+FROM `bfcte_1`

tests/unit/core/compile/sqlglot/expressions/test_unary_compiler.py

@@ -533,6 +533,21 @@ def test_str_contains_regex(scalar_types_df: bpd.DataFrame, snapshot):
     snapshot.assert_match(sql, "out.sql")
 
 
+def test_str_find(scalar_types_df: bpd.DataFrame, snapshot):
+    bf_df = scalar_types_df[["string_col"]]
+    sql = _apply_unary_op(bf_df, ops.StrFindOp("e", start=0, end=None), "string_col")
+    snapshot.assert_match(sql, "out.sql")
+
+    sql = _apply_unary_op(bf_df, ops.StrFindOp("e", start=2, end=None), "string_col")
+    snapshot.assert_match(sql, "out_with_start.sql")
+
+    sql = _apply_unary_op(bf_df, ops.StrFindOp("e", start=0, end=5), "string_col")
+    snapshot.assert_match(sql, "out_with_end.sql")
+
+    sql = _apply_unary_op(bf_df, ops.StrFindOp("e", start=2, end=5), "string_col")
+    snapshot.assert_match(sql, "out_with_start_and_end.sql")
+
+
 def test_strip(scalar_types_df: bpd.DataFrame, snapshot):
     bf_df = scalar_types_df[["string_col"]]
     sql = _apply_unary_op(bf_df, ops.StrStripOp(" "), "string_col")