Replace custom css parsing code with tinycss2

Author: marknsikora

ci/deps/actions-38-minimum_versions.yaml

@@ -49,6 +49,7 @@ dependencies:
   - scipy=1.7.1
   - sqlalchemy=1.4.16
   - tabulate=0.8.9
+  - tinycss2=1.0.0
   - tzdata=2022a
   - xarray=0.19.0
   - xlrd=2.0.1

doc/source/getting_started/install.rst

@@ -296,6 +296,7 @@ Dependency                Minimum Version    Notes
 matplotlib                3.3.2              Plotting library
 Jinja2                    3.0.0              Conditional formatting with DataFrame.style
 tabulate                  0.8.9              Printing in Markdown-friendly format (see `tabulate`_)
+tinycss2                  1.0.0              Style formatting with DataFrame.style
 ========================= ================== =============================================================
 
 Computation

doc/source/whatsnew/v1.6.0.rst

@@ -29,6 +29,7 @@ enhancement2
 Other enhancements
 ^^^^^^^^^^^^^^^^^^
 - :meth:`Series.add_suffix`, :meth:`DataFrame.add_suffix`, :meth:`Series.add_prefix` and :meth:`DataFrame.add_prefix` support an ``axis`` argument. If ``axis`` is set, the default behaviour of which axis to consider can be overwritten (:issue:`47819`)
+- Replace custom css parser with tinycss2 library (:issue:`48868`)
 -
 
 .. ---------------------------------------------------------------------------

pandas/compat/_optional.py

@@ -42,6 +42,7 @@
     "sqlalchemy": "1.4.16",
     "tables": "3.6.1",
     "tabulate": "0.8.9",
+    "tinycss2": "1.0.0",
     "xarray": "0.19.0",
     "xlrd": "2.0.1",
     "xlwt": "1.3.0",

pandas/io/formats/css.py

@@ -13,9 +13,12 @@
 )
 import warnings
 
+from pandas.compat._optional import import_optional_dependency
 from pandas.errors import CSSWarning
 from pandas.util._exceptions import find_stack_level
 
+tinycss2 = import_optional_dependency("tinycss2")
+
 
 def _side_expander(prop_fmt: str) -> Callable:
     """
@@ -377,8 +380,20 @@ def _error():
 
     def atomize(self, declarations: Iterable) -> Generator[tuple[str, str], None, None]:
         for prop, value in declarations:
-            prop = prop.lower()
-            value = value.lower()
+            # Need to reparse the value here in case the prop was passed
+            # through as a dict from the styler rather than through this
+            # classes parse method
+            tokens = []
+            for token in tinycss2.parse_component_value_list(value):
+                if token.type == "ident":
+                    # The old css parser normalized all identifier values, do
+                    # so here to keep backwards compatibility
+                    token.value = token.lower_value
+
+                tokens.append(token)
+
+            value = tinycss2.serialize(tokens).strip()
+
             if prop in self.CSS_EXPANSIONS:
                 expand = self.CSS_EXPANSIONS[prop]
                 yield from expand(self, prop, value)
@@ -395,18 +410,29 @@ def parse(self, declarations_str: str) -> Iterator[tuple[str, str]]:
         ----------
         declarations_str : str
         """
-        for decl in declarations_str.split(";"):
-            if not decl.strip():
-                continue
-            prop, sep, val = decl.partition(":")
-            prop = prop.strip().lower()
-            # TODO: don't lowercase case sensitive parts of values (strings)
-            val = val.strip().lower()
-            if sep:
-                yield prop, val
-            else:
+        declarations = tinycss2.parse_declaration_list(
+            declarations_str, skip_comments=True, skip_whitespace=True
+        )
+
+        for decl in declarations:
+            if decl.type == "error":
                 warnings.warn(
-                    f"Ill-formatted attribute: expected a colon in {repr(decl)}",
+                    decl.message,
                     CSSWarning,
                     stacklevel=find_stack_level(inspect.currentframe()),
                 )
+            else:
+                tokens = []
+                for token in decl.value:
+                    if token.type == "ident":
+                        # The old css parser normalized all identifier values,
+                        # do so here to keep backwards compatibility
+                        token.value = token.lower_value
+
+                    tokens.append(token)
+
+                value = tinycss2.serialize(tokens).strip()
+                if decl.important:
+                    value = f"{value} !important"
+
+                yield decl.lower_name, value

pandas/io/formats/excel.py

@@ -28,6 +28,7 @@
     IndexLabel,
     StorageOptions,
 )
+from pandas.compat._optional import import_optional_dependency
 from pandas.util._decorators import doc
 from pandas.util._exceptions import find_stack_level
 
@@ -54,6 +55,8 @@
 from pandas.io.formats.format import get_level_lengths
 from pandas.io.formats.printing import pprint_thing
 
+tinycss2 = import_optional_dependency("tinycss2")
+
 
 class ExcelCell:
     __fields__ = ("row", "col", "val", "style", "mergestart", "mergeend")
@@ -328,7 +331,24 @@ def build_fill(self, props: Mapping[str, str]):
 
     def build_number_format(self, props: Mapping[str, str]) -> dict[str, str | None]:
         fc = props.get("number-format")
+
+        # Old work around for getting a literal ';'
         fc = fc.replace("§", ";") if isinstance(fc, str) else fc
+
+        # If a quoted string is passed as the value to number format, get the
+        # real value of the string. This will allow passing all the characters
+        # that would otherwise break css if unquoted.
+        if isinstance(fc, str):
+            tokens = tinycss2.parse_component_value_list(fc)
+
+            try:
+                (token,) = tokens
+            except ValueError:
+                pass
+            else:
+                if token.type == "string":
+                    fc = token.value
+
         return {"format_code": fc}
 
     def build_font(

pandas/io/formats/style_render.py

@@ -49,6 +49,8 @@
 jinja2 = import_optional_dependency("jinja2", extra="DataFrame.style requires jinja2.")
 from markupsafe import escape as escape_html  # markupsafe is jinja2 dependency
 
+tinycss2 = import_optional_dependency("tinycss2")
+
 BaseFormatter = Union[str, Callable]
 ExtFormatter = Union[BaseFormatter, Dict[Any, Optional[BaseFormatter]]]
 CSSPair = Tuple[str, Union[str, float]]
@@ -1106,10 +1108,11 @@ def format(
         method to create `to_excel` permissible formatting. Note that semi-colons are
         CSS protected characters but used as separators in Excel's format string.
         Replace semi-colons with the section separator character (ASCII-245) when
-        defining the formatting here.
+        defining the formatting here, or wrap the entire string in quotes to
+        have the value passed directly through to the writter.
 
         >>> df = pd.DataFrame({"A": [1, 0, -1]})
-        >>> pseudo_css = "number-format: 0§[Red](0)§-§@;"
+        >>> pseudo_css = "number-format: '0;[Red](0);-;@';"
         >>> df.style.applymap(lambda v: css).to_excel("formatted_file.xlsx")
         ...  # doctest: +SKIP
 
@@ -1853,18 +1856,26 @@ def maybe_convert_css_to_tuples(style: CSSProperties) -> CSSList:
                                              ('border','1px solid red')]
     """
     if isinstance(style, str):
-        s = style.split(";")
-        try:
-            return [
-                (x.split(":")[0].strip(), x.split(":")[1].strip())
-                for x in s
-                if x.strip() != ""
-            ]
-        except IndexError:
-            raise ValueError(
-                "Styles supplied as string must follow CSS rule formats, "
-                f"for example 'attr: val;'. '{style}' was given."
-            )
+        declarations = tinycss2.parse_declaration_list(
+            style, skip_comments=True, skip_whitespace=True
+        )
+
+        parsed_styles = []
+        for decl in declarations:
+            if decl.type == "error":
+                raise ValueError(
+                    "Styles supplied as string must follow CSS rule formats, "
+                    f"for example 'attr: val;'. '{style}' was given. "
+                    f"{decl.message}"
+                )
+
+            value = tinycss2.serialize(decl.value).strip()
+            if decl.important:
+                value = f"{value} !important"
+
+            parsed_styles.append((decl.name, value))
+
+        return parsed_styles
     return style
 
 

pandas/tests/io/formats/test_to_excel.py

@@ -215,7 +215,11 @@
         ("number-format: 0%", {"number_format": {"format_code": "0%"}}),
         (
             "number-format: 0§[Red](0)§-§@;",
-            {"number_format": {"format_code": "0;[red](0);-;@"}},  # GH 46152
+            {"number_format": {"format_code": "0;[Red](0);-;@"}},  # GH 46152
+        ),
+        (
+            "number-format: '#,##0_);[Red](#,##0)';",
+            {"number_format": {"format_code": "#,##0_);[Red](#,##0)"}},
         ),
     ],
 )