feat(mysql): implement StringToTimestamp

Added test_string_to_timestamp for general backend.

Author: anjakefala

ibis/backends/mysql/registry.py

@@ -112,6 +112,18 @@ def _timestamp_diff(t, expr):
     return sa.func.timestampdiff(sa.text('SECOND'), sa_right, sa_left)
 
 
+def _string_to_timestamp(t, expr):
+    op = expr.op()
+    sa_arg = t.translate(op.arg)
+    sa_format_str = t.translate(op.format_str)
+    if (op.timezone is not None) and op.timezone.op().value != "UTC":
+        raise com.UnsupportedArgumentError(
+            'MySQL backend only supports timezone UTC for converting'
+            'string to timestamp.'
+        )
+    return sa.func.str_to_date(sa_arg, sa_format_str)
+
+
 def _literal(_, expr):
     if isinstance(expr, ir.IntervalScalar):
         if expr.type().unit in {'ms', 'ns'}:
@@ -190,6 +202,7 @@ def _find_in_set(t, expr):
         ops.TimestampAdd: fixed_arity(operator.add, 2),
         ops.TimestampSub: fixed_arity(operator.sub, 2),
         ops.TimestampDiff: _timestamp_diff,
+        ops.StringToTimestamp: _string_to_timestamp,
         ops.DateTruncate: _truncate,
         ops.TimestampTruncate: _truncate,
         ops.IntervalFromInteger: _interval_from_integer,

ibis/backends/tests/test_temporal.py

@@ -10,6 +10,7 @@
 from pytest import param
 
 import ibis
+import ibis.common.exceptions as com
 import ibis.expr.datatypes as dt
 from ibis.backends.pandas.execution.temporal import day_name
 
@@ -563,7 +564,7 @@ def test_strftime(backend, alltypes, df, expr_fn, pandas_pattern):
     ],
 )
 @pytest.mark.notimpl(["datafusion", "mysql", "postgres", "sqlite"])
-def test_to_timestamp(backend, con, unit):
+def test_integer_to_timestamp(backend, con, unit):
     backend_unit = backend.returned_timestamp_unit
     factor = unit_factors[unit]
 
@@ -579,6 +580,75 @@ def test_to_timestamp(backend, con, unit):
     assert result == expected
 
 
+@pytest.mark.parametrize(
+    'fmt, timezone',
+    [
+        # "11/01/10" - "month/day/year"
+        param(
+            '%m/%d/%y',
+            "UTC",
+            id="mysql_format",
+            marks=pytest.mark.never(
+                ["pyspark"], reason="datetime formatting style not supported"
+            ),
+        ),
+        param(
+            'MM/dd/yy',
+            "UTC",
+            id="pyspark_format",
+            marks=pytest.mark.never(
+                ["mysql"], reason="datetime formatting style not supported"
+            ),
+        ),
+    ],
+)
+@pytest.mark.notimpl(
+    [
+        'dask',
+        'pandas',
+        'postgres',
+        'duckdb',
+        'clickhouse',
+        'sqlite',
+        'impala',
+        'datafusion',
+    ]
+)
+def test_string_to_timestamp(backend, con, fmt, timezone):
+    table = con.table('functional_alltypes')
+    result = table.mutate(
+        date=table.date_string_col.to_timestamp(fmt, timezone)
+    ).execute()
+
+    # TEST: do we get the same date out, that we put in?
+    # format string assumes that we are using pandas' strftime
+    for i, val in enumerate(result["date"]):
+        assert val.strftime("%m/%d/%y") == result["date_string_col"][i]
+
+
+@pytest.mark.notimpl(
+    [
+        'dask',
+        'pandas',
+        'postgres',
+        'duckdb',
+        'clickhouse',
+        'sqlite',
+        'impala',
+        'datafusion',
+    ]
+)
+def test_string_to_timestamp_tz_error(backend, con):
+    table = con.table('functional_alltypes')
+
+    with pytest.raises(com.UnsupportedArgumentError):
+        table.mutate(
+            date=table.date_string_col.to_timestamp(
+                "%m/%d/%y", 'non-utc-timezone'
+            )
+        ).compile()
+
+
 @pytest.mark.parametrize(
     ('date', 'expected_index', 'expected_day'),
     [