fakesnow 0.9.36__py3-none-any.whl → 0.9.38__py3-none-any.whl

fakesnow/transforms/transforms.py

@@ -0,0 +1,1335 @@
+from __future__ import annotations
+
+from pathlib import Path
+from string import Template
+from typing import ClassVar, cast
+
+import sqlglot
+from sqlglot import exp
+
+from fakesnow.variables import Variables
+
+SUCCESS_NOP = sqlglot.parse_one("SELECT 'Statement executed successfully.' as status")
+
+
+def alias_in_join(expression: exp.Expression) -> exp.Expression:
+    if (
+        isinstance(expression, exp.Select)
+        and (aliases := {e.args.get("alias"): e for e in expression.expressions if isinstance(e, exp.Alias)})
+        and (joins := expression.args.get("joins"))
+    ):
+        j: exp.Join
+        for j in joins:
+            if (
+                (on := j.args.get("on"))
+                and (col := on.this)
+                and (isinstance(col, exp.Column))
+                and (alias := aliases.get(col.this))
+                # don't rewrite col with table identifier
+                and not col.table
+            ):
+                col.args["this"] = alias.this
+
+    return expression
+
+
+def alter_table_strip_cluster_by(expression: exp.Expression) -> exp.Expression:
+    """Turn alter table cluster by into a no-op"""
+    if (
+        isinstance(expression, exp.Alter)
+        and (actions := expression.args.get("actions"))
+        and len(actions) == 1
+        and (isinstance(actions[0], exp.Cluster))
+    ):
+        return SUCCESS_NOP
+    return expression
+
+
+def array_size(expression: exp.Expression) -> exp.Expression:
+    if isinstance(expression, exp.ArraySize):
+        # return null if not json array
+        jal = exp.Anonymous(this="json_array_length", expressions=[expression.this])
+        is_json_array = exp.EQ(
+            this=exp.Anonymous(this="json_type", expressions=[expression.this]),
+            expression=exp.Literal(this="ARRAY", is_string=True),
+        )
+        return exp.Case(ifs=[exp.If(this=is_json_array, true=jal)])
+
+    return expression
+
+
+def array_agg(expression: exp.Expression) -> exp.Expression:
+    if isinstance(expression, exp.ArrayAgg) and not isinstance(expression.parent, exp.Window):
+        return exp.Anonymous(this="TO_JSON", expressions=[expression])
+
+    if isinstance(expression, exp.Window) and isinstance(expression.this, exp.ArrayAgg):
+        return exp.Anonymous(this="TO_JSON", expressions=[expression])
+
+    return expression
+
+
+def array_agg_within_group(expression: exp.Expression) -> exp.Expression:
+    """Convert ARRAY_AGG(<expr>) WITHIN GROUP (<order-by-clause>) to ARRAY_AGG( <expr> <order-by-clause> )
+    Snowflake uses ARRAY_AGG(<expr>) WITHIN GROUP (ORDER BY <order-by-clause>)
+    to order the array, but DuckDB uses ARRAY_AGG( <expr> <order-by-clause> ).
+    See;
+        - https://docs.snowflake.com/en/sql-reference/functions/array_agg
+        - https://duckdb.org/docs/sql/aggregates.html#order-by-clause-in-aggregate-functions
+    Note; Snowflake has following restriction;
+            If you specify DISTINCT and WITHIN GROUP, both must refer to the same column.
+          Transformation does not handle this restriction.
+    """
+    if (
+        isinstance(expression, exp.WithinGroup)
+        and (agg := expression.find(exp.ArrayAgg))
+        and (order := expression.expression)
+    ):
+        return exp.ArrayAgg(
+            this=exp.Order(
+                this=agg.this,
+                expressions=order.expressions,
+            )
+        )
+
+    return expression
+
+
+def create_clone(expression: exp.Expression) -> exp.Expression:
+    """Transform create table clone to create table as select."""
+
+    if (
+        isinstance(expression, exp.Create)
+        and str(expression.args.get("kind")).upper() == "TABLE"
+        and (clone := expression.find(exp.Clone))
+    ):
+        return exp.Create(
+            this=expression.this,
+            kind="TABLE",
+            expression=exp.Select(
+                expressions=[
+                    exp.Star(),
+                ],
+                **{"from": exp.From(this=clone.this)},
+            ),
+        )
+    return expression
+
+
+# TODO: move this into a Dialect as a transpilation
+def create_database(expression: exp.Expression, db_path: Path | None = None) -> exp.Expression:
+    """Transform create database to attach database.
+
+    Example:
+        >>> import sqlglot
+        >>> sqlglot.parse_one("CREATE database foo").transform(create_database).sql()
+        'ATTACH DATABASE ':memory:' as foo'
+    Args:
+        expression (exp.Expression): the expression that will be transformed.
+
+    Returns:
+        exp.Expression: The transformed expression, with the database name stored in the create_db_name arg.
+    """
+
+    if isinstance(expression, exp.Create) and str(expression.args.get("kind")).upper() == "DATABASE":
+        ident = expression.find(exp.Identifier)
+        assert ident, f"No identifier in {expression.sql}"
+        db_name = ident.this
+        db_file = f"{db_path / db_name}.db" if db_path else ":memory:"
+
+        if_not_exists = "IF NOT EXISTS " if expression.args.get("exists") else ""
+
+        return exp.Command(
+            this="ATTACH",
+            expression=exp.Literal(this=f"{if_not_exists}DATABASE '{db_file}' AS {db_name}", is_string=True),
+            create_db_name=db_name,
+        )
+
+    return expression
+
+
+SQL_DESCRIBE_TABLE = Template(
+    """
+SELECT
+    column_name AS "name",
+    CASE WHEN data_type = 'NUMBER' THEN 'NUMBER(' || numeric_precision || ',' || numeric_scale || ')'
+         WHEN data_type = 'TEXT' THEN 'VARCHAR(' || coalesce(character_maximum_length,16777216)  || ')'
+         WHEN data_type = 'TIMESTAMP_NTZ' THEN 'TIMESTAMP_NTZ(9)'
+         WHEN data_type = 'TIMESTAMP_TZ' THEN 'TIMESTAMP_TZ(9)'
+         WHEN data_type = 'TIME' THEN 'TIME(9)'
+         WHEN data_type = 'BINARY' THEN 'BINARY(8388608)'
+        ELSE data_type END AS "type",
+    'COLUMN' AS "kind",
+    CASE WHEN is_nullable = 'YES' THEN 'Y' ELSE 'N' END AS "null?",
+    column_default AS "default",
+    'N' AS "primary key",
+    'N' AS "unique key",
+    NULL::VARCHAR AS "check",
+    NULL::VARCHAR AS "expression",
+    NULL::VARCHAR AS "comment",
+    NULL::VARCHAR AS "policy name",
+    NULL::JSON AS "privacy domain",
+FROM _fs_information_schema._fs_columns
+WHERE table_catalog = '${catalog}' AND table_schema = '${schema}' AND table_name = '${table}'
+ORDER BY ordinal_position
+"""
+)
+
+SQL_DESCRIBE_INFO_SCHEMA = Template(
+    """
+SELECT
+    column_name AS "name",
+    column_type as "type",
+    'COLUMN' AS "kind",
+    CASE WHEN "null" = 'YES' THEN 'Y' ELSE 'N' END AS "null?",
+    NULL::VARCHAR AS "default",
+    'N' AS "primary key",
+    'N' AS "unique key",
+    NULL::VARCHAR AS "check",
+    NULL::VARCHAR AS "expression",
+    NULL::VARCHAR AS "comment",
+    NULL::VARCHAR AS "policy name",
+    NULL::JSON AS "privacy domain",
+FROM (DESCRIBE ${view})
+"""
+)
+
+
+def describe_table(
+    expression: exp.Expression, current_database: str | None = None, current_schema: str | None = None
+) -> exp.Expression:
+    """Redirect to the information_schema._fs_columns to match snowflake.
+
+    See https://docs.snowflake.com/en/sql-reference/sql/desc-table
+    """
+
+    if (
+        isinstance(expression, exp.Describe)
+        and (kind := expression.args.get("kind"))
+        and isinstance(kind, str)
+        and kind.upper() in ("TABLE", "VIEW")
+        and (table := expression.find(exp.Table))
+    ):
+        catalog = table.catalog or current_database
+        schema = table.db or current_schema
+
+        if schema and schema.upper() == "_FS_INFORMATION_SCHEMA":
+            # describing an information_schema view
+            # (schema already transformed from information_schema -> _fs_information_schema)
+            return sqlglot.parse_one(SQL_DESCRIBE_INFO_SCHEMA.substitute(view=f"{schema}.{table.name}"), read="duckdb")
+
+        return sqlglot.parse_one(
+            SQL_DESCRIBE_TABLE.substitute(catalog=catalog, schema=schema, table=table.name),
+            read="duckdb",
+        )
+
+    return expression
+
+
+def drop_schema_cascade(expression: exp.Expression) -> exp.Expression:  #
+    """Drop schema cascade.
+
+    By default duckdb won't delete a schema if it contains tables, whereas snowflake will.
+    So we add the cascade keyword to mimic snowflake's behaviour.
+
+    Example:
+        >>> import sqlglot
+        >>> sqlglot.parse_one("DROP SCHEMA schema1").transform(remove_comment).sql()
+        'DROP SCHEMA schema1 cascade'
+    Args:
+        expression (exp.Expression): the expression that will be transformed.
+
+    Returns:
+        exp.Expression: The transformed expression.
+    """
+
+    if (
+        not isinstance(expression, exp.Drop)
+        or not (kind := expression.args.get("kind"))
+        or not isinstance(kind, str)
+        or kind.upper() != "SCHEMA"
+    ):
+        return expression
+
+    new = expression.copy()
+    new.args["cascade"] = True
+    return new
+
+
+def dateadd_date_cast(expression: exp.Expression) -> exp.Expression:
+    """Cast result of DATEADD to DATE if the given expression is a cast to DATE
+       and unit is either DAY, WEEK, MONTH or YEAR to mimic Snowflake's DATEADD
+       behaviour.
+
+    Snowflake;
+        SELECT DATEADD(DAY, 3, '2023-03-03'::DATE) as D;
+            D: 2023-03-06 (DATE)
+    DuckDB;
+        SELECT CAST('2023-03-03' AS DATE) + INTERVAL 3 DAY AS D
+            D: 2023-03-06 00:00:00 (TIMESTAMP)
+    """
+
+    if not isinstance(expression, exp.DateAdd):
+        return expression
+
+    if expression.unit is None:
+        return expression
+
+    if not isinstance(expression.unit.this, str):
+        return expression
+
+    if (unit := expression.unit.this.upper()) and unit.upper() not in {"DAY", "WEEK", "MONTH", "YEAR"}:
+        return expression
+
+    if not isinstance(expression.this, exp.Cast):
+        return expression
+
+    if expression.this.to.this != exp.DataType.Type.DATE:
+        return expression
+
+    return exp.Cast(
+        this=expression,
+        to=exp.DataType(this=exp.DataType.Type.DATE, nested=False, prefix=False),
+    )
+
+
+def dateadd_string_literal_timestamp_cast(expression: exp.Expression) -> exp.Expression:
+    """Snowflake's DATEADD function implicitly casts string literals to
+    timestamps regardless of unit.
+    """
+    if not isinstance(expression, exp.DateAdd):
+        return expression
+
+    if not isinstance(expression.this, exp.Literal) or not expression.this.is_string:
+        return expression
+
+    new_dateadd = expression.copy()
+    new_dateadd.set(
+        "this",
+        exp.Cast(
+            this=expression.this,
+            # TODO: support TIMESTAMP_TYPE_MAPPING of TIMESTAMP_LTZ/TZ
+            to=exp.DataType(this=exp.DataType.Type.TIMESTAMP, nested=False, prefix=False),
+        ),
+    )
+
+    return new_dateadd
+
+
+def datediff_string_literal_timestamp_cast(expression: exp.Expression) -> exp.Expression:
+    """Snowflake's DATEDIFF function implicitly casts string literals to
+    timestamps regardless of unit.
+    """
+
+    if not isinstance(expression, exp.DateDiff):
+        return expression
+
+    op1 = expression.this.copy()
+    op2 = expression.expression.copy()
+
+    if isinstance(op1, exp.Literal) and op1.is_string:
+        op1 = exp.Cast(
+            this=op1,
+            # TODO: support TIMESTAMP_TYPE_MAPPING of TIMESTAMP_LTZ/TZ
+            to=exp.DataType(this=exp.DataType.Type.TIMESTAMP, nested=False, prefix=False),
+        )
+
+    if isinstance(op2, exp.Literal) and op2.is_string:
+        op2 = exp.Cast(
+            this=op2,
+            # TODO: support TIMESTAMP_TYPE_MAPPING of TIMESTAMP_LTZ/TZ
+            to=exp.DataType(this=exp.DataType.Type.TIMESTAMP, nested=False, prefix=False),
+        )
+
+    new_datediff = expression.copy()
+    new_datediff.set("this", op1)
+    new_datediff.set("expression", op2)
+
+    return new_datediff
+
+
+def extract_comment_on_columns(expression: exp.Expression) -> exp.Expression:
+    """Extract column comments, removing it from the Expression.
+
+    duckdb doesn't support comments. So we remove them from the expression and store them in the column_comment arg.
+    We also replace the transform the expression to NOP if the statement can't be executed by duckdb.
+
+    Args:
+        expression (exp.Expression): the expression that will be transformed.
+
+    Returns:
+        exp.Expression: The transformed expression, with any comment stored in the new 'table_comment' arg.
+    """
+
+    if isinstance(expression, exp.Alter) and (actions := expression.args.get("actions")):
+        new_actions: list[exp.Expression] = []
+        col_comments: list[tuple[str, str]] = []
+        for a in actions:
+            if isinstance(a, exp.AlterColumn) and (comment := a.args.get("comment")):
+                col_comments.append((a.name, comment.this))
+            else:
+                new_actions.append(a)
+        if not new_actions:
+            expression = SUCCESS_NOP.copy()
+        else:
+            expression.set("actions", new_actions)
+        expression.args["col_comments"] = col_comments
+
+    return expression
+
+
+def extract_comment_on_table(expression: exp.Expression) -> exp.Expression:
+    """Extract table comment, removing it from the Expression.
+
+    duckdb doesn't support comments. So we remove them from the expression and store them in the table_comment arg.
+    We also replace the transform the expression to NOP if the statement can't be executed by duckdb.
+
+    Args:
+        expression (exp.Expression): the expression that will be transformed.
+
+    Returns:
+        exp.Expression: The transformed expression, with any comment stored in the new 'table_comment' arg.
+    """
+
+    if isinstance(expression, exp.Create) and (table := expression.find(exp.Table)):
+        comment = None
+        if props := cast(exp.Properties, expression.args.get("properties")):
+            other_props = []
+            for p in props.expressions:
+                if isinstance(p, exp.SchemaCommentProperty) and (isinstance(p.this, (exp.Literal, exp.Var))):
+                    comment = p.this.this
+                else:
+                    other_props.append(p)
+
+            new = expression.copy()
+            new_props: exp.Properties = new.args["properties"]
+            new_props.set("expressions", other_props)
+            new.args["table_comment"] = (table, comment)
+            return new
+    elif (
+        isinstance(expression, exp.Comment)
+        and (cexp := expression.args.get("expression"))
+        and (table := expression.find(exp.Table))
+    ):
+        new = SUCCESS_NOP.copy()
+        new.args["table_comment"] = (table, cexp.this)
+        return new
+    elif (
+        isinstance(expression, exp.Alter)
+        and (sexp := expression.find(exp.AlterSet))
+        and (scp := sexp.find(exp.SchemaCommentProperty))
+        and isinstance(scp.this, exp.Literal)
+        and (table := expression.find(exp.Table))
+    ):
+        new = SUCCESS_NOP.copy()
+        new.args["table_comment"] = (table, scp.this.this)
+        return new
+
+    return expression
+
+
+def extract_text_length(expression: exp.Expression) -> exp.Expression:
+    """Extract length of text columns.
+
+    duckdb doesn't have fixed-sized text types. So we capture the size of text types and store that in the
+    character_maximum_length arg.
+
+    Args:
+        expression (exp.Expression): the expression that will be transformed.
+
+    Returns:
+        exp.Expression: The original expression, with any text lengths stored in the new 'text_lengths' arg.
+    """
+
+    if isinstance(expression, (exp.Create, exp.Alter)):
+        text_lengths = []
+
+        # exp.Select is for a ctas, exp.Schema is a plain definition
+        if cols := expression.find(exp.Select, exp.Schema):
+            expressions = cols.expressions
+        else:
+            # alter table
+            expressions = expression.args.get("actions") or []
+        for e in expressions:
+            if dts := [
+                dt for dt in e.find_all(exp.DataType) if dt.this in (exp.DataType.Type.VARCHAR, exp.DataType.Type.TEXT)
+            ]:
+                col_name = e.alias if isinstance(e, exp.Alias) else e.name
+                if len(dts) == 1 and (dt_size := dts[0].find(exp.DataTypeParam)):
+                    size = (
+                        isinstance(dt_size.this, exp.Literal)
+                        and isinstance(dt_size.this.this, str)
+                        and int(dt_size.this.this)
+                    )
+                else:
+                    size = 16777216
+                text_lengths.append((col_name, size))
+
+        if text_lengths:
+            expression.args["text_lengths"] = text_lengths
+
+    return expression
+
+
+def flatten(expression: exp.Expression) -> exp.Expression:
+    """Flatten an array.
+
+    See https://docs.snowflake.com/en/sql-reference/functions/flatten
+
+    TODO: support objects.
+    """
+    if (isinstance(expression, (exp.Lateral, exp.TableFromRows))) and isinstance(expression.this, exp.Explode):
+        input_ = (
+            expression.this.this.expression if isinstance(expression.this.this, exp.Kwarg) else expression.this.this
+        )
+        alias = expression.args.get("alias")
+        return exp.Table(this=exp.Anonymous(this="_fs_flatten", expressions=[input_]), alias=alias)
+
+    return expression
+
+
+def flatten_value_cast_as_varchar(expression: exp.Expression) -> exp.Expression:
+    """Return raw unquoted string when flatten VALUE is cast to varchar.
+
+    Returns a raw string using the Duckdb ->> operator, aka the json_extract_string function, see
+    https://duckdb.org/docs/extensions/json#json-extraction-functions
+    """
+    if (
+        isinstance(expression, exp.Cast)
+        and isinstance(expression.this, exp.Column)
+        and expression.this.name.upper() == "VALUE"
+        and expression.to.this in [exp.DataType.Type.VARCHAR, exp.DataType.Type.TEXT]
+        and (select := expression.find_ancestor(exp.Select))
+        and select.find(exp.Explode)
+    ):
+        return exp.JSONExtractScalar(this=expression.this, expression=exp.JSONPath(expressions=[exp.JSONPathRoot()]))
+
+    return expression
+
+
+def float_to_double(expression: exp.Expression) -> exp.Expression:
+    """Convert float to double for 64 bit precision.
+
+    Snowflake floats are all 64 bit (ie: double)
+    see https://docs.snowflake.com/en/sql-reference/data-types-numeric#float-float4-float8
+    """
+
+    if isinstance(expression, exp.DataType) and expression.this == exp.DataType.Type.FLOAT:
+        expression.args["this"] = exp.DataType.Type.DOUBLE
+
+    return expression
+
+
+def identifier(expression: exp.Expression) -> exp.Expression:
+    """Convert identifier function to an identifier.
+
+    See https://docs.snowflake.com/en/sql-reference/identifier-literal
+    """
+
+    if (
+        isinstance(expression, exp.Anonymous)
+        and isinstance(expression.this, str)
+        and expression.this.upper() == "IDENTIFIER"
+    ):
+        expression = exp.Identifier(this=expression.expressions[0].this, quoted=False)
+
+    return expression
+
+
+def indices_to_json_extract(expression: exp.Expression) -> exp.Expression:
+    """Convert indices on objects and arrays to json_extract or json_extract_string
+
+    Supports Snowflake array indices, see
+    https://docs.snowflake.com/en/sql-reference/data-types-semistructured#accessing-elements-of-an-array-by-index-or-by-slice
+    and object indices, see
+    https://docs.snowflake.com/en/sql-reference/data-types-semistructured#accessing-elements-of-an-object-by-key
+
+    Duckdb uses the -> operator, aka the json_extract function, see
+    https://duckdb.org/docs/extensions/json#json-extraction-functions
+
+    This works for Snowflake arrays too because we convert them to JSON in duckdb.
+    """
+    if (
+        isinstance(expression, exp.Bracket)
+        and len(expression.expressions) == 1
+        and (index := expression.expressions[0])
+        and isinstance(index, exp.Literal)
+        and index.this
+    ):
+        if isinstance(expression.parent, exp.Cast) and expression.parent.to.this == exp.DataType.Type.VARCHAR:
+            # If the parent is a cast to varchar, we need to use JSONExtractScalar
+            # to get the unquoted string value.
+            klass = exp.JSONExtractScalar
+        else:
+            klass = exp.JSONExtract
+        if index.is_string:
+            return klass(this=expression.this, expression=exp.Literal(this=f"$.{index.this}", is_string=True))
+        else:
+            return klass(this=expression.this, expression=exp.Literal(this=f"$[{index.this}]", is_string=True))
+
+    return expression
+
+
+def information_schema_fs_columns(expression: exp.Expression) -> exp.Expression:
+    """Redirect to the _FS_COLUMNS view which has metadata that matches snowflake.
+
+    Because duckdb doesn't store character_maximum_length or character_octet_length.
+    """
+
+    if (
+        isinstance(expression, exp.Table)
+        and expression.db
+        and expression.db.upper() == "INFORMATION_SCHEMA"
+        and expression.name
+        and expression.name.upper() == "COLUMNS"
+    ):
+        expression.set("this", exp.Identifier(this="_FS_COLUMNS", quoted=False))
+        expression.set("db", exp.Identifier(this="_FS_INFORMATION_SCHEMA", quoted=False))
+
+    return expression
+
+
+def information_schema_databases(
+    expression: exp.Expression,
+    current_schema: str | None = None,
+) -> exp.Expression:
+    if (
+        isinstance(expression, exp.Table)
+        and (
+            expression.db.upper() == "INFORMATION_SCHEMA"
+            or (current_schema and current_schema.upper() == "INFORMATION_SCHEMA")
+        )
+        and expression.name.upper() == "DATABASES"
+    ):
+        return exp.Table(
+            this=exp.Identifier(this="DATABASES", quoted=False),
+            db=exp.Identifier(this="_FS_INFORMATION_SCHEMA", quoted=False),
+        )
+    return expression
+
+
+def information_schema_fs_tables(
+    expression: exp.Expression,
+) -> exp.Expression:
+    """Use _FS_TABLES to access additional metadata columns (eg: comment)."""
+
+    if (
+        isinstance(expression, exp.Select)
+        and (tbl := expression.find(exp.Table))
+        and tbl.db.upper() == "INFORMATION_SCHEMA"
+        and tbl.name.upper() == "TABLES"
+    ):
+        tbl.set("this", exp.Identifier(this="_FS_TABLES", quoted=False))
+        tbl.set("db", exp.Identifier(this="_FS_INFORMATION_SCHEMA", quoted=False))
+
+    return expression
+
+
+def information_schema_fs_views(expression: exp.Expression) -> exp.Expression:
+    """Use _FS_VIEWS to return Snowflake's version instead of duckdb's."""
+
+    if (
+        isinstance(expression, exp.Select)
+        and (tbl := expression.find(exp.Table))
+        and tbl.db.upper() == "INFORMATION_SCHEMA"
+        and tbl.name.upper() == "VIEWS"
+    ):
+        tbl.set("this", exp.Identifier(this="_FS_VIEWS", quoted=False))
+        tbl.set("db", exp.Identifier(this="_FS_INFORMATION_SCHEMA", quoted=False))
+
+    return expression
+
+
+NUMBER_38_0 = [
+    exp.DataTypeParam(this=exp.Literal(this="38", is_string=False)),
+    exp.DataTypeParam(this=exp.Literal(this="0", is_string=False)),
+]
+
+
+def integer_precision(expression: exp.Expression) -> exp.Expression:
+    """Convert integers and number(38,0) to bigint.
+
+    So fetch_all will return int and dataframes will return them with a dtype of int64.
+    """
+    if (
+        isinstance(expression, exp.DataType)
+        and expression.this == exp.DataType.Type.DECIMAL
+        and (not expression.expressions or expression.expressions == NUMBER_38_0)
+    ) or expression.this in (exp.DataType.Type.INT, exp.DataType.Type.SMALLINT, exp.DataType.Type.TINYINT):
+        return exp.DataType(
+            this=exp.DataType.Type.BIGINT,
+            nested=False,
+            prefix=False,
+        )
+
+    return expression
+
+
+def json_extract_cased_as_varchar(expression: exp.Expression) -> exp.Expression:
+    """Convert json to varchar inside JSONExtract.
+
+    Snowflake case conversion (upper/lower) turns variant into varchar. This
+    mimics that behaviour within get_path.
+
+    TODO: a generic version that works on any variant, not just JSONExtract
+
+    Returns a raw string using the Duckdb ->> operator, aka the json_extract_string function, see
+    https://duckdb.org/docs/extensions/json#json-extraction-functions
+    """
+    if (
+        isinstance(expression, (exp.Upper, exp.Lower))
+        and (gp := expression.this)
+        and isinstance(gp, exp.JSONExtract)
+        and (path := gp.expression)
+        and isinstance(path, exp.JSONPath)
+    ):
+        expression.set("this", exp.JSONExtractScalar(this=gp.this, expression=path))
+
+    return expression
+
+
+def json_extract_cast_as_varchar(expression: exp.Expression) -> exp.Expression:
+    """Return raw unquoted string when casting json extraction to varchar.
+
+    Returns a raw string using the Duckdb ->> operator, aka the json_extract_string function, see
+    https://duckdb.org/docs/extensions/json#json-extraction-functions
+    """
+    if (
+        isinstance(expression, exp.Cast)
+        and (je := expression.this)
+        and isinstance(je, exp.JSONExtract)
+        and (path := je.expression)
+        and isinstance(path, exp.JSONPath)
+    ):
+        je.replace(exp.JSONExtractScalar(this=je.this, expression=path))
+    return expression
+
+
+def json_extract_precedence(expression: exp.Expression) -> exp.Expression:
+    """Associate json extract operands to avoid duckdb operators of higher precedence transforming the expression.
+
+    See https://github.com/tekumara/fakesnow/issues/53
+    """
+    if isinstance(expression, (exp.JSONExtract, exp.JSONExtractScalar)):
+        return exp.Paren(this=expression)
+    return expression
+
+
+def random(expression: exp.Expression) -> exp.Expression:
+    """Convert random() and random(seed).
+
+    Snowflake random() is an signed 64 bit integer.
+    Duckdb random() is a double between 0 and 1 and uses setseed() to set the seed.
+    """
+    if isinstance(expression, exp.Select) and (rand := expression.find(exp.Rand)):
+        # shift result to between min and max signed 64bit integer
+        new_rand = exp.Cast(
+            this=exp.Paren(
+                this=exp.Mul(
+                    this=exp.Paren(this=exp.Sub(this=exp.Rand(), expression=exp.Literal(this="0.5", is_string=False))),
+                    expression=exp.Literal(this="9223372036854775807", is_string=False),
+                )
+            ),
+            to=exp.DataType(this=exp.DataType.Type.BIGINT, nested=False, prefix=False),
+        )
+
+        rand.replace(new_rand)
+
+        # convert seed to double between 0 and 1 by dividing by max INTEGER (int32)
+        # (not max BIGINT (int64) because we don't have enough floating point precision to distinguish seeds)
+        # then attach to SELECT as the seed arg
+        # (we can't attach it to exp.Rand because it will be rendered in the sql)
+        if rand.this and isinstance(rand.this, exp.Literal):
+            expression.args["seed"] = f"{rand.this}/2147483647-0.5"
+
+    return expression
+
+
+def sample(expression: exp.Expression) -> exp.Expression:
+    if isinstance(expression, exp.TableSample) and not expression.args.get("method"):
+        # set snowflake default (bernoulli) rather than use the duckdb default (system)
+        # because bernoulli works better at small row sizes like we have in tests
+        expression.set("method", exp.Var(this="BERNOULLI"))
+
+    return expression
+
+
+def object_construct(expression: exp.Expression) -> exp.Expression:
+    """Convert OBJECT_CONSTRUCT to TO_JSON.
+
+    Internally snowflake stores OBJECT types as a json string, so the Duckdb JSON type most closely matches.
+
+    See https://docs.snowflake.com/en/sql-reference/functions/object_construct
+    """
+
+    if not isinstance(expression, exp.Struct):
+        return expression
+
+    non_null_expressions = []
+    for e in expression.expressions:
+        if not (isinstance(e, exp.PropertyEQ)):
+            non_null_expressions.append(e)
+            continue
+
+        left = e.left
+        right = e.right
+
+        left_is_null = isinstance(left, exp.Null)
+        right_is_null = isinstance(right, exp.Null)
+
+        if left_is_null or right_is_null:
+            continue
+
+        non_null_expressions.append(e)
+
+    new_struct = expression.copy()
+    new_struct.set("expressions", non_null_expressions)
+    return exp.Anonymous(this="TO_JSON", expressions=[new_struct])
+
+
+def regex_replace(expression: exp.Expression) -> exp.Expression:
+    """Transform regex_replace expressions from snowflake to duckdb."""
+
+    if isinstance(expression, exp.RegexpReplace) and isinstance(expression.expression, exp.Literal):
+        if len(expression.args) > 3:
+            # see https://docs.snowflake.com/en/sql-reference/functions/regexp_replace
+            raise NotImplementedError(
+                "REGEXP_REPLACE with additional parameters (eg: <position>, <occurrence>, <parameters>)"
+            )
+
+        # pattern: snowflake requires escaping backslashes in single-quoted string constants, but duckdb doesn't
+        # see https://docs.snowflake.com/en/sql-reference/functions-regexp#label-regexp-escape-character-caveats
+        expression.args["expression"] = exp.Literal(
+            this=expression.expression.this.replace("\\\\", "\\"), is_string=True
+        )
+
+        if not expression.args.get("replacement"):
+            # if no replacement string, the snowflake default is ''
+            expression.args["replacement"] = exp.Literal(this="", is_string=True)
+
+        # snowflake regex replacements are global
+        expression.args["modifiers"] = exp.Literal(this="g", is_string=True)
+
+    return expression
+
+
+def regex_substr(expression: exp.Expression) -> exp.Expression:
+    """Transform regex_substr expressions from snowflake to duckdb.
+
+    See https://docs.snowflake.com/en/sql-reference/functions/regexp_substr
+    """
+
+    if isinstance(expression, exp.RegexpExtract):
+        subject = expression.this
+
+        # pattern: snowflake requires escaping backslashes in single-quoted string constants, but duckdb doesn't
+        # see https://docs.snowflake.com/en/sql-reference/functions-regexp#label-regexp-escape-character-caveats
+        pattern = expression.expression
+        pattern.args["this"] = pattern.this.replace("\\\\", "\\")
+
+        # number of characters from the beginning of the string where the function starts searching for matches
+        position = expression.args["position"] or exp.Literal(this="1", is_string=False)
+
+        # which occurrence of the pattern to match
+        occurrence = expression.args["occurrence"]
+        occurrence = int(occurrence.this) if occurrence else 1
+
+        # the duckdb dialect increments bracket (ie: index) expressions by 1 because duckdb is 1-indexed,
+        # so we need to compensate by subtracting 1
+        occurrence = exp.Literal(this=str(occurrence - 1), is_string=False)
+
+        if parameters := expression.args["parameters"]:
+            # 'e' parameter doesn't make sense for duckdb
+            regex_parameters = exp.Literal(this=parameters.this.replace("e", ""), is_string=True)
+        else:
+            regex_parameters = exp.Literal(is_string=True)
+
+        group_num = expression.args["group"]
+        if not group_num:
+            if isinstance(regex_parameters.this, str) and "e" in regex_parameters.this:
+                group_num = exp.Literal(this="1", is_string=False)
+            else:
+                group_num = exp.Literal(this="0", is_string=False)
+
+        expression = exp.Bracket(
+            this=exp.Anonymous(
+                this="regexp_extract_all",
+                expressions=[
+                    # slice subject from position onwards
+                    exp.Bracket(this=subject, expressions=[exp.Slice(this=position)]),
+                    pattern,
+                    group_num,
+                    regex_parameters,
+                ],
+            ),
+            # select index of occurrence
+            expressions=[occurrence],
+        )
+
+    return expression
+
+
+# TODO: move this into a Dialect as a transpilation
+def set_schema(expression: exp.Expression, current_database: str | None) -> exp.Expression:
+    """Transform USE SCHEMA/DATABASE to SET schema.
+
+    Example:
+        >>> import sqlglot
+        >>> sqlglot.parse_one("USE SCHEMA bar").transform(set_schema, current_database="foo").sql()
+        "SET schema = 'foo.bar'"
+        >>> sqlglot.parse_one("USE SCHEMA foo.bar").transform(set_schema).sql()
+        "SET schema = 'foo.bar'"
+        >>> sqlglot.parse_one("USE DATABASE marts").transform(set_schema).sql()
+        "SET schema = 'marts.main'"
+
+        See tests for more examples.
+    Args:
+        expression (exp.Expression): the expression that will be transformed.
+
+    Returns:
+        exp.Expression: A SET schema expression if the input is a USE
+            expression, otherwise expression is returned as-is.
+    """
+
+    if (
+        isinstance(expression, exp.Use)
+        and (kind := expression.args.get("kind"))
+        and isinstance(kind, exp.Var)
+        and kind.name
+        and kind.name.upper() in ["SCHEMA", "DATABASE"]
+    ):
+        assert expression.this, f"No identifier for USE expression {expression}"
+
+        if kind.name.upper() == "DATABASE":
+            # duckdb's default schema is main
+            database = expression.this.name
+            return exp.Command(
+                this="SET", expression=exp.Literal.string(f"schema = '{database}.main'"), set_database=database
+            )
+        else:
+            # SCHEMA
+            if db := expression.this.args.get("db"):  # noqa: SIM108
+                db_name = db.name
+            else:
+                # isn't qualified with a database
+                db_name = current_database
+
+            # assertion always true because check_db_schema is called before this
+            assert db_name
+
+            schema = expression.this.name
+            return exp.Command(
+                this="SET", expression=exp.Literal.string(f"schema = '{db_name}.{schema}'"), set_schema=schema
+            )
+
+    return expression
+
+
+def split(expression: exp.Expression) -> exp.Expression:
+    """
+    Convert output of duckdb str_split from varchar[] to JSON array to match Snowflake.
+    """
+    if isinstance(expression, exp.Split):
+        return exp.Anonymous(this="to_json", expressions=[expression])
+
+    return expression
+
+
+def tag(expression: exp.Expression) -> exp.Expression:
+    """Handle tags. Transfer tags into upserts of the tag table.
+
+    duckdb doesn't support tags. In lieu of a full implementation, for now we make it a NOP.
+
+    Example:
+        >>> import sqlglot
+        >>> sqlglot.parse_one("ALTER TABLE table1 SET TAG foo='bar'").transform(tag).sql()
+        "SELECT 'Statement executed successfully.'"
+    Args:
+        expression (exp.Expression): the expression that will be transformed.
+
+    Returns:
+        exp.Expression: The transformed expression.
+    """
+
+    if isinstance(expression, exp.Alter) and (actions := expression.args.get("actions")):
+        for a in actions:
+            if isinstance(a, exp.AlterSet) and a.args.get("tag"):
+                return SUCCESS_NOP
+    elif (
+        isinstance(expression, exp.Command)
+        and (cexp := expression.args.get("expression"))
+        and isinstance(cexp, str)
+        and "SET TAG" in cexp.upper()
+    ):
+        # alter table modify column set tag
+        return SUCCESS_NOP
+    elif (
+        isinstance(expression, exp.Create)
+        and (kind := expression.args.get("kind"))
+        and isinstance(kind, str)
+        and kind.upper() == "TAG"
+    ):
+        return SUCCESS_NOP
+
+    return expression
+
+
+def to_date(expression: exp.Expression) -> exp.Expression:
+    """Convert to_date() to a cast.
+
+    See https://docs.snowflake.com/en/sql-reference/functions/to_date
+
+    Example:
+        >>> import sqlglot
+        >>> sqlglot.parse_one("SELECT to_date(to_timestamp(0))").transform(to_date).sql()
+        "SELECT CAST(DATE_TRUNC('day', TO_TIMESTAMP(0)) AS DATE)"
+    Args:
+        expression (exp.Expression): the expression that will be transformed.
+
+    Returns:
+        exp.Expression: The transformed expression.
+    """
+
+    if (
+        isinstance(expression, exp.Anonymous)
+        and isinstance(expression.this, str)
+        and expression.this.upper() == "TO_DATE"
+    ):
+        return exp.Cast(
+            this=expression.expressions[0],
+            to=exp.DataType(this=exp.DataType.Type.DATE, nested=False, prefix=False),
+        )
+    return expression
+
+
+def _get_to_number_args(e: exp.ToNumber) -> tuple[exp.Expression | None, exp.Expression | None, exp.Expression | None]:
+    arg_format = e.args.get("format")
+    arg_precision = e.args.get("precision")
+    arg_scale = e.args.get("scale")
+
+    _format = None
+    _precision = None
+    _scale = None
+
+    # to_number(value, <format>, <precision>, <scale>)
+    if arg_format:
+        if arg_format.is_string:
+            # to_number('100', 'TM9' ...)
+            _format = arg_format
+
+            # to_number('100', 'TM9', 10 ...)
+            if arg_precision:
+                _precision = arg_precision
+
+                # to_number('100', 'TM9', 10, 2)
+                if arg_scale:
+                    _scale = arg_scale
+        else:
+            # to_number('100', 10, ...)
+            # arg_format is not a string, so it must be precision.
+            _precision = arg_format
+
+            # to_number('100', 10, 2)
+            # And arg_precision must be scale
+            if arg_precision:
+                _scale = arg_precision
+    elif arg_precision:
+        _precision = arg_precision
+        if arg_scale:
+            _scale = arg_scale
+
+    return _format, _precision, _scale
+
+
+def _to_decimal(expression: exp.Expression, cast_node: type[exp.Cast]) -> exp.Expression:
+    expressions: list[exp.Expression] = expression.expressions
+
+    if len(expressions) > 1 and expressions[1].is_string:
+        # see https://docs.snowflake.com/en/sql-reference/functions/to_decimal#arguments
+        raise NotImplementedError(f"{expression.this} with format argument")
+
+    precision = expressions[1] if len(expressions) > 1 else exp.Literal(this="38", is_string=False)
+    scale = expressions[2] if len(expressions) > 2 else exp.Literal(this="0", is_string=False)
+
+    return cast_node(
+        this=expressions[0],
+        to=exp.DataType(this=exp.DataType.Type.DECIMAL, expressions=[precision, scale], nested=False, prefix=False),
+    )
+
+
+def to_decimal(expression: exp.Expression) -> exp.Expression:
+    """Transform to_decimal, to_number, to_numeric expressions from snowflake to duckdb.
+
+    See https://docs.snowflake.com/en/sql-reference/functions/to_decimal
+    """
+
+    if isinstance(expression, exp.ToNumber):
+        format_, precision, scale = _get_to_number_args(expression)
+        if format_:
+            raise NotImplementedError(f"{expression.this} with format argument")
+
+        if not precision:
+            precision = exp.Literal(this="38", is_string=False)
+        if not scale:
+            scale = exp.Literal(this="0", is_string=False)
+
+        return exp.Cast(
+            this=expression.this,
+            to=exp.DataType(this=exp.DataType.Type.DECIMAL, expressions=[precision, scale], nested=False, prefix=False),
+        )
+
+    if (
+        isinstance(expression, exp.Anonymous)
+        and isinstance(expression.this, str)
+        and expression.this.upper() in ["TO_DECIMAL", "TO_NUMERIC"]
+    ):
+        return _to_decimal(expression, exp.Cast)
+
+    return expression
+
+
+def try_to_decimal(expression: exp.Expression) -> exp.Expression:
+    """Transform try_to_decimal, try_to_number, try_to_numeric expressions from snowflake to duckdb.
+    See https://docs.snowflake.com/en/sql-reference/functions/try_to_decimal
+    """
+
+    if (
+        isinstance(expression, exp.Anonymous)
+        and isinstance(expression.this, str)
+        and expression.this.upper() in ["TRY_TO_DECIMAL", "TRY_TO_NUMBER", "TRY_TO_NUMERIC"]
+    ):
+        return _to_decimal(expression, exp.TryCast)
+
+    return expression
+
+
+def to_timestamp(expression: exp.Expression) -> exp.Expression:
+    """Convert to_timestamp(seconds) to timestamp without timezone (ie: TIMESTAMP_NTZ).
+
+    See https://docs.snowflake.com/en/sql-reference/functions/to_timestamp
+    """
+
+    if isinstance(expression, exp.UnixToTime):
+        return exp.Cast(
+            this=expression,
+            to=exp.DataType(this=exp.DataType.Type.TIMESTAMP, nested=False, prefix=False),
+        )
+    return expression
+
+
+def to_timestamp_ntz(expression: exp.Expression) -> exp.Expression:
+    """Convert to_timestamp_ntz to to_timestamp (StrToTime).
+
+    Because it's not yet supported by sqlglot, see https://github.com/tobymao/sqlglot/issues/2748
+    """
+
+    if isinstance(expression, exp.Anonymous) and (
+        isinstance(expression.this, str) and expression.this.upper() == "TO_TIMESTAMP_NTZ"
+    ):
+        return exp.StrToTime(
+            this=expression.expressions[0],
+            format=exp.Literal(this="%Y-%m-%d %H:%M:%S", is_string=True),
+        )
+    return expression
+
+
+def timestamp_ntz(expression: exp.Expression) -> exp.Expression:
+    """Convert timestamp_ntz (snowflake) to timestamp (duckdb).
+
+    NB: timestamp_ntz defaults to nanosecond precision (ie: NTZ(9)). The duckdb equivalent is TIMESTAMP_NS.
+    However we use TIMESTAMP (ie: microsecond precision) here rather than TIMESTAMP_NS to avoid
+    https://github.com/duckdb/duckdb/issues/7980 in test_write_pandas_timestamp_ntz.
+    """
+
+    if isinstance(expression, exp.DataType) and expression.this == exp.DataType.Type.TIMESTAMPNTZ:
+        return exp.DataType(this=exp.DataType.Type.TIMESTAMP)
+
+    return expression
+
+
+def trim_cast_varchar(expression: exp.Expression) -> exp.Expression:
+    """Snowflake's TRIM casts input to VARCHAR implicitly."""
+
+    if not (isinstance(expression, exp.Trim)):
+        return expression
+
+    operand = expression.this
+    if isinstance(operand, exp.Cast) and operand.to.this in [exp.DataType.Type.VARCHAR, exp.DataType.Type.TEXT]:
+        return expression
+
+    return exp.Trim(
+        this=exp.Cast(this=operand, to=exp.DataType(this=exp.DataType.Type.VARCHAR, nested=False, prefix=False))
+    )
+
+
+def try_parse_json(expression: exp.Expression) -> exp.Expression:
+    """Convert TRY_PARSE_JSON() to TRY_CAST(... as JSON).
+
+    Example:
+        >>> import sqlglot
+        >>> sqlglot.parse_one("select try_parse_json('{}')").transform(parse_json).sql()
+        "SELECT TRY_CAST('{}' AS JSON)"
+    Args:
+        expression (exp.Expression): the expression that will be transformed.
+
+    Returns:
+        exp.Expression: The transformed expression.
+    """
+
+    if (
+        isinstance(expression, exp.Anonymous)
+        and isinstance(expression.this, str)
+        and expression.this.upper() == "TRY_PARSE_JSON"
+    ):
+        expressions = expression.expressions
+        return exp.TryCast(
+            this=expressions[0],
+            to=exp.DataType(this=exp.DataType.Type.JSON, nested=False),
+        )
+
+    return expression
+
+
+def semi_structured_types(expression: exp.Expression) -> exp.Expression:
+    """Convert OBJECT, ARRAY, and VARIANT types to duckdb compatible types.
+
+    Example:
+        >>> import sqlglot
+        >>> sqlglot.parse_one("CREATE TABLE table1 (name object)").transform(semi_structured_types).sql()
+        "CREATE TABLE table1 (name JSON)"
+    Args:
+        expression (exp.Expression): the expression that will be transformed.
+
+    Returns:
+        exp.Expression: The transformed expression.
+    """
+
+    if isinstance(expression, exp.DataType) and expression.this in [
+        exp.DataType.Type.ARRAY,
+        exp.DataType.Type.OBJECT,
+        exp.DataType.Type.VARIANT,
+    ]:
+        new = expression.copy()
+        new.args["this"] = exp.DataType.Type.JSON
+        return new
+
+    return expression
+
+
+def upper_case_unquoted_identifiers(expression: exp.Expression) -> exp.Expression:
+    """Upper case unquoted identifiers.
+
+    Snowflake represents case-insensitivity using upper-case identifiers in cursor results.
+    duckdb uses lowercase. We convert all unquoted identifiers to uppercase to match snowflake.
+
+    Example:
+        >>> import sqlglot
+        >>> sqlglot.parse_one("select name, name as fname from table1").transform(upper_case_unquoted_identifiers).sql()
+        'SELECT NAME, NAME AS FNAME FROM TABLE1'
+    Args:
+        expression (exp.Expression): the expression that will be transformed.
+
+    Returns:
+        exp.Expression: The transformed expression.
+    """
+
+    if isinstance(expression, exp.Identifier) and not expression.quoted and isinstance(expression.this, str):
+        new = expression.copy()
+        new.set("this", expression.this.upper())
+        return new
+
+    return expression
+
+
+def values_columns(expression: exp.Expression) -> exp.Expression:
+    """Support column1, column2 expressions in VALUES.
+
+    Snowflake uses column1, column2 .. for unnamed columns in VALUES. Whereas duckdb uses col0, col1 ..
+    See https://docs.snowflake.com/en/sql-reference/constructs/values#examples
+    """
+
+    if (
+        isinstance(expression, exp.Values)
+        and not expression.alias
+        and expression.find_ancestor(exp.Select)
+        and (values := expression.find(exp.Tuple))
+    ):
+        num_columns = len(values.expressions)
+        columns = [exp.Identifier(this=f"COLUMN{i + 1}", quoted=True) for i in range(num_columns)]
+        expression.set("alias", exp.TableAlias(this=exp.Identifier(this="_", quoted=False), columns=columns))
+
+    return expression
+
+
+def create_user(expression: exp.Expression) -> exp.Expression:
+    """Transform CREATE USER to a query against the global database's information_schema._fs_users table.
+
+    https://docs.snowflake.com/en/sql-reference/sql/create-user
+    """
+    # XXX: this is a placeholder. We need to implement the full CREATE USER syntax, but
+    #      sqlglot doesnt yet support Create for snowflake.
+    if isinstance(expression, exp.Command) and expression.this == "CREATE":
+        sub_exp = expression.expression.strip()
+        if sub_exp.upper().startswith("USER"):
+            _, name, *ignored = sub_exp.split(" ")
+            if ignored:
+                raise NotImplementedError(f"`CREATE USER` with {ignored}")
+            return sqlglot.parse_one(
+                f"INSERT INTO _fs_global._fs_information_schema._fs_users (name) VALUES ('{name}')", read="duckdb"
+            )
+
+    return expression
+
+
+def update_variables(
+    expression: exp.Expression,
+    variables: Variables,
+) -> exp.Expression:
+    if Variables.is_variable_modifier(expression):
+        variables.update_variables(expression)
+        return SUCCESS_NOP  # Nothing further to do if its a SET/UNSET operation.
+    return expression
+
+
+class SHA256(exp.Func):
+    _sql_names: ClassVar = ["SHA256"]
+    arg_types: ClassVar = {"this": True}
+
+
+def sha256(expression: exp.Expression) -> exp.Expression:
+    """Convert sha2() or sha2_hex() to sha256().
+
+    Convert sha2_binary() to unhex(sha256()).
+
+    Example:
+        >>> import sqlglot
+        >>> sqlglot.parse_one("insert into table1 (name) select sha2('foo')").transform(sha256).sql()
+        "INSERT INTO table1 (name) SELECT SHA256('foo')"
+    Args:
+        expression (exp.Expression): the expression that will be transformed.
+
+    Returns:
+        exp.Expression: The transformed expression.
+    """
+
+    if isinstance(expression, exp.SHA2) and expression.args.get("length", exp.Literal.number(256)).this == "256":
+        return SHA256(this=expression.this)
+    elif (
+        isinstance(expression, exp.Anonymous)
+        and expression.this.upper() == "SHA2_HEX"
+        and (
+            len(expression.expressions) == 1
+            or (len(expression.expressions) == 2 and expression.expressions[1].this == "256")
+        )
+    ):
+        return SHA256(this=expression.expressions[0])
+    elif (
+        isinstance(expression, exp.Anonymous)
+        and expression.this.upper() == "SHA2_BINARY"
+        and (
+            len(expression.expressions) == 1
+            or (len(expression.expressions) == 2 and expression.expressions[1].this == "256")
+        )
+    ):
+        return exp.Unhex(this=SHA256(this=expression.expressions[0]))
+
+    return expression