fix(ir): make impure ibis.random() and ibis.uuid() functions return unique node instances (ibis-project#8967)

Co-authored-by: Phillip Cloud <417981+cpcloud@users.noreply.github.com>

Author: kszucs

ibis/backends/bigquery/compiler.py

@@ -120,8 +120,6 @@ class BigQueryCompiler(SQLGlotCompiler):
         ops.RPad: "rpad",
         ops.Levenshtein: "edit_distance",
         ops.Modulus: "mod",
-        ops.RandomScalar: "rand",
-        ops.RandomUUID: "generate_uuid",
         ops.RegexReplace: "regexp_replace",
         ops.RegexSearch: "regexp_contains",
         ops.Time: "time",
@@ -695,3 +693,6 @@ def visit_CountDistinct(self, op, *, arg, where):
         if where is not None:
             arg = self.if_(where, arg, NULL)
         return self.f.count(sge.Distinct(expressions=[arg]))
+
+    def visit_RandomUUID(self, op, **kwargs):
+        return self.f.generate_uuid()

ibis/backends/clickhouse/compiler.py

@@ -98,8 +98,6 @@ class ClickHouseCompiler(SQLGlotCompiler):
         ops.NotNull: "isNotNull",
         ops.NullIf: "nullIf",
         ops.RStrip: "trimRight",
-        ops.RandomScalar: "randCanonical",
-        ops.RandomUUID: "generateUUIDv4",
         ops.RegexReplace: "replaceRegexpAll",
         ops.RowNumber: "row_number",
         ops.StartsWith: "startsWith",
@@ -635,6 +633,12 @@ def visit_TimestampRange(self, op, *, start, stop, step):
     def visit_RegexSplit(self, op, *, arg, pattern):
         return self.f.splitByRegexp(pattern, self.cast(arg, dt.String(nullable=False)))
 
+    def visit_RandomScalar(self, op, **kwargs):
+        return self.f.randCanonical()
+
+    def visit_RandomUUID(self, op, **kwargs):
+        return self.f.generateUUIDv4()
+
     @staticmethod
     def _generate_groups(groups):
         return groups

ibis/backends/datafusion/compiler.py

@@ -75,7 +75,6 @@ class DataFusionCompiler(SQLGlotCompiler):
         ops.Last: "last_value",
         ops.Median: "median",
         ops.StringLength: "character_length",
-        ops.RandomUUID: "uuid",
         ops.RegexSplit: "regex_split",
         ops.EndsWith: "ends_with",
         ops.ArrayIntersect: "array_intersect",

ibis/backends/druid/compiler.py

@@ -59,6 +59,7 @@ class DruidCompiler(SQLGlotCompiler):
             ops.Median,
             ops.MultiQuantile,
             ops.Quantile,
+            ops.RandomUUID,
             ops.RegexReplace,
             ops.RegexSplit,
             ops.RowID,

ibis/backends/duckdb/compiler.py

@@ -48,7 +48,6 @@ class DuckDBCompiler(SQLGlotCompiler):
         ops.MapMerge: "map_concat",
         ops.MapValues: "map_values",
         ops.Mode: "mode",
-        ops.RandomUUID: "uuid",
         ops.TimeFromHMS: "make_time",
         ops.TypeOf: "typeof",
         ops.GeoPoint: "st_point",
@@ -450,3 +449,9 @@ def visit_StructField(self, op, *, arg, field):
                 expression=sg.to_identifier(field, quoted=self.quoted),
             )
         return super().visit_StructField(op, arg=arg, field=field)
+
+    def visit_RandomScalar(self, op, **kwargs):
+        return self.f.random()
+
+    def visit_RandomUUID(self, op, **kwargs):
+        return self.f.uuid()

ibis/backends/exasol/compiler.py

@@ -67,6 +67,7 @@ class ExasolCompiler(SQLGlotCompiler):
             ops.Median,
             ops.MultiQuantile,
             ops.Quantile,
+            ops.RandomUUID,
             ops.ReductionVectorizedUDF,
             ops.RegexExtract,
             ops.RegexReplace,

ibis/backends/flink/compiler.py

@@ -79,8 +79,6 @@ class FlinkCompiler(SQLGlotCompiler):
         ops.MapKeys: "map_keys",
         ops.MapValues: "map_values",
         ops.Power: "power",
-        ops.RandomScalar: "rand",
-        ops.RandomUUID: "uuid",
         ops.RegexSearch: "regexp",
         ops.StrRight: "right",
         ops.StringLength: "char_length",

ibis/backends/impala/compiler.py

@@ -76,7 +76,6 @@ class ImpalaCompiler(SQLGlotCompiler):
         ops.Hash: "fnv_hash",
         ops.LStrip: "ltrim",
         ops.Ln: "ln",
-        ops.RandomUUID: "uuid",
         ops.RStrip: "rtrim",
         ops.Strip: "trim",
         ops.TypeOf: "typeof",
@@ -146,7 +145,7 @@ def visit_CountDistinct(self, op, *, arg, where):
     def visit_Xor(self, op, *, left, right):
         return sg.and_(sg.or_(left, right), sg.not_(sg.and_(left, right)))
 
-    def visit_RandomScalar(self, op):
+    def visit_RandomScalar(self, op, **_):
         return self.f.rand(self.f.utc_to_unix_micros(self.f.utc_timestamp()))
 
     def visit_DayOfWeekIndex(self, op, *, arg):

ibis/backends/mssql/compiler.py

@@ -129,8 +129,6 @@ class MSSQLCompiler(SQLGlotCompiler):
         ops.Ln: "log",
         ops.Log10: "log10",
         ops.Power: "power",
-        ops.RandomScalar: "rand",
-        ops.RandomUUID: "newid",
         ops.Repeat: "replicate",
         ops.Reverse: "reverse",
         ops.StringAscii: "ascii",
@@ -172,6 +170,9 @@ def _minimize_spec(start, end, spec):
             return None
         return spec
 
+    def visit_RandomUUID(self, op, **kwargs):
+        return self.f.newid()
+
     def visit_StringLength(self, op, *, arg):
         """The MSSQL LEN function doesn't count trailing spaces.
 

ibis/backends/oracle/compiler.py

@@ -80,6 +80,7 @@ class OracleCompiler(SQLGlotCompiler):
             ops.ExtractWeekOfYear,
             ops.ExtractDayOfYear,
             ops.RowID,
+            ops.RandomUUID,
         )
     )
 
@@ -221,7 +222,7 @@ def visit_Log(self, op, *, arg, base):
     def visit_IsInf(self, op, *, arg):
         return arg.isin(self.POS_INF, self.NEG_INF)
 
-    def visit_RandomScalar(self, op):
+    def visit_RandomScalar(self, op, **_):
         # Not using FuncGen here because of dotted function call
         return sg.func("dbms_random.value")