refactor(padding): follow python string padding conventions (ibis-pro…

…ject#10096)

BREAKING CHANGE: String padding operations now follow Python semantics and leave strings greater than the padding length untouched.

ibis/backends/impala/tests/snapshots/test_string_builtins/test_string_builtins/lpad_char/out.sql

@@ -1,3 +1,3 @@
 SELECT
-  LPAD(`t0`.`string_col`, 1, 'a') AS `LPad(string_col, 1, 'a')`
+  LPAD(`t0`.`string_col`, GREATEST(LENGTH(`t0`.`string_col`), 1), 'a') AS `LPad(string_col, 1, 'a')`
 FROM `functional_alltypes` AS `t0`

ibis/backends/impala/tests/snapshots/test_string_builtins/test_string_builtins/lpad_default/out.sql

@@ -1,3 +1,3 @@
 SELECT
-  LPAD(`t0`.`string_col`, 25, ' ') AS `LPad(string_col, 25, ' ')`
+  LPAD(`t0`.`string_col`, GREATEST(LENGTH(`t0`.`string_col`), 25), ' ') AS `LPad(string_col, 25, ' ')`
 FROM `functional_alltypes` AS `t0`

ibis/backends/impala/tests/snapshots/test_string_builtins/test_string_builtins/rpad_char/out.sql

@@ -1,3 +1,3 @@
 SELECT
-  RPAD(`t0`.`string_col`, 1, 'a') AS `RPad(string_col, 1, 'a')`
+  RPAD(`t0`.`string_col`, GREATEST(LENGTH(`t0`.`string_col`), 1), 'a') AS `RPad(string_col, 1, 'a')`
 FROM `functional_alltypes` AS `t0`

ibis/backends/impala/tests/snapshots/test_string_builtins/test_string_builtins/rpad_default/out.sql

@@ -1,3 +1,3 @@
 SELECT
-  RPAD(`t0`.`string_col`, 25, ' ') AS `RPad(string_col, 25, ' ')`
+  RPAD(`t0`.`string_col`, GREATEST(LENGTH(`t0`.`string_col`), 25), ' ') AS `RPad(string_col, 25, ' ')`
 FROM `functional_alltypes` AS `t0`

ibis/backends/impala/tests/test_exprs.py

@@ -295,9 +295,9 @@ def test_decimal_builtins_2(con, func, expected):
         (L("0123").translate("012", "abc"), "abc3"),
         (L("abcd").find("a"), 0),
         (L("baaaab").find("b", 2), 5),
-        (L("abcd").lpad(1, "-"), "a"),
+        (L("abcd").lpad(1, "-"), "abcd"),
         (L("abcd").lpad(5), " abcd"),
-        (L("abcd").rpad(1, "-"), "a"),
+        (L("abcd").rpad(1, "-"), "abcd"),
         (L("abcd").rpad(5), "abcd "),
         (L("abcd").find_in_set(["a", "b", "abcd"]), 2),
         (L(", ").join(["a", "b"]), "a, b"),

ibis/backends/sql/compilers/base.py

@@ -331,7 +331,6 @@ class SQLGlotCompiler(abc.ABC):
         ops.IsInf: "isinf",
         ops.IsNan: "isnan",
         ops.JSONGetItem: "json_extract",
-        ops.LPad: "lpad",
         LastValue: "last_value",
         ops.Levenshtein: "levenshtein",
         ops.Ln: "ln",
@@ -347,7 +346,6 @@ class SQLGlotCompiler(abc.ABC):
         ops.PercentRank: "percent_rank",
         ops.Pi: "pi",
         ops.Power: "pow",
-        ops.RPad: "rpad",
         ops.Radians: "radians",
         ops.RegexSearch: "regexp_like",
         ops.RegexSplit: "regexp_split",
@@ -985,6 +983,12 @@ def visit_RStrip(self, op, *, arg):
     def visit_LStrip(self, op, *, arg):
         return self.f.ltrim(arg, string.whitespace)
 
+    def visit_LPad(self, op, *, arg, length, pad):
+        return self.f.lpad(arg, self.f.greatest(self.f.length(arg), length), pad)
+
+    def visit_RPad(self, op, *, arg, length, pad):
+        return self.f.rpad(arg, self.f.greatest(self.f.length(arg), length), pad)
+
     def visit_Substring(self, op, *, arg, start, length):
         if isinstance(op.length, ops.Literal) and (value := op.length.value) < 0:
             raise com.IbisInputError(

ibis/backends/sql/compilers/bigquery/__init__.py

@@ -183,8 +183,6 @@ class BigQueryCompiler(SQLGlotCompiler):
         ops.IsInf: "is_inf",
         ops.IsNan: "is_nan",
         ops.Log10: "log10",
-        ops.LPad: "lpad",
-        ops.RPad: "rpad",
         ops.Levenshtein: "edit_distance",
         ops.Modulus: "mod",
         ops.RegexReplace: "regexp_replace",

ibis/backends/sql/compilers/clickhouse.py

@@ -480,6 +480,16 @@ def visit_Strip(self, op, *, arg):
             this=arg, position="BOTH", expression=sge.Literal.string(whitespace)
         )
 
+    def visit_LPad(self, op, *, arg, length, pad):
+        return self.f.leftPadUTF8(
+            arg, self.f.greatest(self.f.lengthUTF8(arg), length), pad
+        )
+
+    def visit_RPad(self, op, *, arg, length, pad):
+        return self.f.rightPadUTF8(
+            arg, self.f.greatest(self.f.lengthUTF8(arg), length), pad
+        )
+
     def visit_DayOfWeekIndex(self, op, *, arg):
         weekdays = len(calendar.day_name)
         return (((self.f.toDayOfWeek(arg) - 1) % weekdays) + weekdays) % weekdays

ibis/backends/sql/compilers/datafusion.py

@@ -207,6 +207,20 @@ def visit_RegexSearch(self, op, *, arg, pattern):
     def visit_StringContains(self, op, *, haystack, needle):
         return self.f.strpos(haystack, needle) > sg.exp.convert(0)
 
+    def visit_LPad(self, op, *, arg, length, pad):
+        return self.if_(
+            length <= self.f.length(arg),
+            arg,
+            self.f.concat(self.f.repeat(pad, length - self.f.length(arg)), arg),
+        )
+
+    def visit_RPad(self, op, *, arg, length, pad):
+        return self.if_(
+            length <= self.f.length(arg),
+            arg,
+            self.f.concat(arg, self.f.repeat(pad, length - self.f.length(arg))),
+        )
+
     def visit_ExtractFragment(self, op, *, arg):
         return self.f.extract_url_field(arg, "fragment")
 

ibis/backends/sql/compilers/duckdb.py

@@ -568,6 +568,20 @@ def visit_Hash(self, op, *, arg):
     def visit_StringConcat(self, op, *, arg):
         return reduce(lambda x, y: sge.DPipe(this=x, expression=y), arg)
 
+    def visit_LPad(self, op, *, arg, length, pad):
+        return self.if_(
+            length <= self.f.length(arg),
+            arg,
+            self.f.concat(self.f.repeat(pad, length - self.f.length(arg)), arg),
+        )
+
+    def visit_RPad(self, op, *, arg, length, pad):
+        return self.if_(
+            length <= self.f.length(arg),
+            arg,
+            self.f.concat(arg, self.f.repeat(pad, length - self.f.length(arg))),
+        )
+
     def visit_StringSlice(self, op, *, arg, start, end):
         if start is not None:
             start += 1

ibis/backends/sql/compilers/flink.py

@@ -518,6 +518,20 @@ def visit_StringFind(self, op, *, arg, substr, start, end):
 
         return self.f.instr(arg, substr)
 
+    def visit_LPad(self, op, *, arg, length, pad):
+        return self.if_(
+            length <= self.f.length(arg),
+            arg,
+            self.f.concat(self.f.repeat(pad, length - self.f.length(arg)), arg),
+        )
+
+    def visit_RPad(self, op, *, arg, length, pad):
+        return self.if_(
+            length <= self.f.length(arg),
+            arg,
+            self.f.concat(arg, self.f.repeat(pad, length - self.f.length(arg))),
+        )
+
     def visit_StartsWith(self, op, *, arg, start):
         return self.f.left(arg, self.f.char_length(start)).eq(start)
 

ibis/backends/sql/compilers/mssql.py

@@ -528,7 +528,7 @@ def visit_LPad(self, op, *, arg, length, pad):
         return self.if_(
             length <= self.f.length(arg),
             arg,
-            self.f.left(
+            self.f.right(
                 self.f.concat(self.f.replicate(pad, length - self.f.length(arg)), arg),
                 length,
             ),

ibis/backends/sql/compilers/oracle.py

@@ -82,8 +82,6 @@ class OracleCompiler(SQLGlotCompiler):
         ops.BitXor: "bit_xor_agg",
         ops.BitwiseAnd: "bitand",
         ops.Hash: "ora_hash",
-        ops.LPad: "lpad",
-        ops.RPad: "rpad",
         ops.StringAscii: "ascii",
         ops.Mode: "stats_mode",
     }
@@ -275,6 +273,12 @@ def visit_StringContains(self, op, *, haystack, needle):
     def visit_StringJoin(self, op, *, arg, sep):
         return self.f.concat(*toolz.interpose(sep, arg))
 
+    def visit_LPad(self, op, *, arg, length, pad):
+        return self.f.lpad(arg, self.f.greatest(self.f.length(arg), length), pad)
+
+    def visit_RPad(self, op, *, arg, length, pad):
+        return self.f.rpad(arg, self.f.greatest(self.f.length(arg), length), pad)
+
     ## Aggregate stuff
 
     def visit_Correlation(self, op, *, left, right, where, how):

ibis/backends/sqlite/udf.py

@@ -258,12 +258,12 @@ def _ibis_string_ascii(string):
 
 @udf
 def _ibis_rpad(string, width, pad):
-    return string.ljust(width, pad)[:width]
+    return string.ljust(width, pad)
 
 
 @udf
 def _ibis_lpad(string, width, pad):
-    return string.rjust(width, pad)[:width]
+    return string.rjust(width, pad)
 
 
 @udf

ibis/backends/tests/sql/snapshots/test_sql/test_no_cart_join/out.sql

@@ -19,9 +19,22 @@ FROM (
         "t1"."ancestor_level_number",
         "t1"."ancestor_node_sort_order",
         "t1"."descendant_node_natural_key",
-        LPAD('-', (
-          "t1"."ancestor_level_number" - 1
-        ) * 7, '-') || "t1"."ancestor_level_name" AS "product_level_name"
+        CASE
+          WHEN (
+            (
+              "t1"."ancestor_level_number" - 1
+            ) * 7
+          ) <= LENGTH('-')
+          THEN '-'
+          ELSE CONCAT(
+            REPEAT('-', (
+              (
+                "t1"."ancestor_level_number" - 1
+              ) * 7
+            ) - LENGTH('-')),
+            '-'
+          )
+        END || "t1"."ancestor_level_name" AS "product_level_name"
       FROM "products" AS "t1"
     ) AS "t4"
       ON "t2"."product_id" = "t4"."descendant_node_natural_key"