Update sqlglot to support 27.0.0 or later (#4227)

## Changes

This PR updates the `sqlglot` requirement to allow the 27.0.0 release.
This release includes support for placeholders in SQL, a relatively new
syntax feature supported in DBR 15.2 (or Apache Spark 4.0).

SQL queries can be of the form:
```sql
SELECT 1 AS col1, 2 AS col2, 3 AS col3 FROM {sdf_system_columns} LIMIT 5
```
…where `sdf_system_columns` is passed as a keyword argument, for
example:
```python
sdf_system_columns = spark.read.table("system.information_schema.columns")
sdf_example = spark.sql("SELECT 1 AS col1, 2 AS col2, 3 AS col3 FROM {sdf_system_columns} LIMIT 5",
                        sdf_system_columns = sdf_system_columns)
```

Older versions of sqlgplot would fail on this with a parsing error, but
now it's parsed correctly but returns a `Placeholder` AST node as the
name of the table instead of a string. Given the complexity of handling
this properly, for now I've chosen to preserve the old behaviour: the
linter will mark the query as unsupported. (See
[es-1285042.py](https://github.com/databrickslabs/ucx/blob/e48281a67fbd9f9b80688a8fd08c0e6e19dab5c6/tests/unit/source_code/samples/functional/es-1285042.py)
for an example of this.)

### Linked issues

Closes #4203 

### Tests

- existing integration tests
- existing (+ 1 updated) unit tests

Author: asnare

pyproject.toml

@@ -49,7 +49,7 @@ dependencies = ["databricks-sdk>=0.58.0,<0.59.0",
                 "databricks-labs-lsql>=0.16.0,<0.17.0",
                 "databricks-labs-blueprint>=0.11.0,<0.12.0",
                 "PyYAML>=6.0.0,<6.1.0",
-                "sqlglot>=26.7.0,<26.8.0",
+                "sqlglot>=26.7.0,<27.1.0",
                 "astroid>=3.3.0,<3.4.0"]
 
 [project.optional-dependencies]

src/databricks/labs/ucx/source_code/sql/sql_parser.py

@@ -3,8 +3,8 @@
 from typing import TypeVar
 
 from sqlglot import parse
-from sqlglot.errors import SqlglotError
-from sqlglot.expressions import Create, Delete, Drop, Expression, Select, Table, Update, Use
+from sqlglot.errors import SqlglotError, UnsupportedError
+from sqlglot.expressions import Create, Delete, Drop, Expression, Select, Table, Update, Use, Identifier
 
 from databricks.labs.ucx.source_code.base import UsedTable, CurrentSessionState
 
@@ -54,6 +54,13 @@ def _collect_table_info(
         if not src_schema:
             logger.warning(f"Could not determine schema for table {table.name}")
             return None
+        # Sqlglot handlers parameter markers by returning an Identifier as the name instead of a string.
+        # For example: {foo} -> Identifier(this=foo)
+        if isinstance(table.name, Identifier):
+            # TODO: Support these properly, for example by inferring the placeholder value from the outside context.
+            msg = f"Table placeholder detected, not yet supported: {{{table.name}}}"
+            logger.debug(msg)
+            raise UnsupportedError(msg)
         return UsedTable(
             catalog_name=catalog_name,
             schema_name=src_schema,