ENH: Styler.to_latex(): conditional styling with native latex format

doc/source/reference/style.rst

@@ -24,6 +24,7 @@ Styler properties
 
    Styler.env
    Styler.template_html
+   Styler.template_latex
    Styler.loader
 
 Style application
@@ -66,3 +67,4 @@ Style export and import
    Styler.export
    Styler.use
    Styler.to_excel
+   Styler.to_latex

doc/source/whatsnew/v1.3.0.rst

@@ -141,6 +141,9 @@ properly format HTML and eliminate some inconsistencies (:issue:`39942` :issue:`
 :class:`.Styler` has also been compatible with non-unique index or columns, at least for as many features as are fully compatible, others made only partially compatible (:issue:`41269`).
 One also has greater control of the display through separate sparsification of the index or columns, using the new 'styler' options context (:issue:`41142`).
 
+We have added an extension to allow LaTeX styling as an alternative to CSS styling and a method :meth:`.Styler.to_latex`
+which renders the necessary LaTeX format including built-up styles.
+
 Documentation has also seen major revisions in light of new features (:issue:`39720` :issue:`39317` :issue:`40493`)
 
 .. _whatsnew_130.dataframe_honors_copy_with_dict:

pandas/io/formats/style.py

@@ -21,6 +21,7 @@
 
 from pandas._typing import (
     Axis,
+    FilePathOrBuffer,
     FrameOrSeries,
     FrameOrSeriesUnion,
     IndexLabel,
@@ -30,6 +31,7 @@
 from pandas.util._decorators import doc
 
 import pandas as pd
+from pandas import RangeIndex
 from pandas.api.types import is_list_like
 from pandas.core import generic
 import pandas.core.common as com
@@ -39,6 +41,8 @@
 )
 from pandas.core.generic import NDFrame
 
+from pandas.io.formats.format import save_to_buffer
+
 jinja2 = import_optional_dependency("jinja2", extra="DataFrame.style requires jinja2.")
 
 from pandas.io.formats.style_render import (
@@ -403,6 +407,338 @@ def to_excel(
             engine=engine,
         )
 
+    def to_latex(
+        self,
+        buf: FilePathOrBuffer[str] | None = None,
+        *,
+        column_format: str | None = None,
+        position: str | None = None,
+        position_float: str | None = None,
+        hrules: bool = False,
+        label: str | None = None,
+        caption: str | None = None,
+        sparse_index: bool | None = None,
+        sparse_columns: bool | None = None,
+        multirow_align: str = "c",
+        multicol_align: str = "r",
+        siunitx: bool = False,
+        encoding: str | None = None,
+    ):
+        r"""
+        Write Styler to a file, buffer or string in LaTeX format.
+
+        .. versionadded:: 1.3.0
+
+        Parameters
+        ----------
+        buf : str, Path, or StringIO-like, optional, default None
+            Buffer to write to. If ``None``, the output is returned as a string.
+        column_format : str, optional
+            The LaTeX column specification placed in location:
+
+            \\begin{tabular}{<column_format>}
+
+            Defaults to 'l' for index and
+            non-numeric data columns, and, for numeric data columns,
+            to 'r' by default, or 'S' if ``siunitx`` is ``True``.
+        position : str, optional
+            The LaTeX positional argument (e.g. 'h!') for tables, placed in location:
+
+            \\begin{table}[<position>]
+        position_float : {"centering", "raggedleft", "raggedright"}, optional
+            The LaTeX float command placed in location:
+
+            \\begin{table}[<position>]
+
+            \\<position_float>
+        hrules : bool, default False
+            Set to `True` to add \\toprule, \\midrule and \\bottomrule from the
+            {booktabs} LaTeX package.
+        label : str, optional
+            The LaTeX label included as: \\label{<label>}.
+            This is used with \\ref{<label>} in the main .tex file.
+        caption : str, optional
+            The LaTeX table caption included as: \\caption{<caption>}.
+        sparse_index : bool, optional
+            Whether to sparsify the display of a hierarchical index. Setting to False
+            will display each explicit level element in a hierarchical key for each row.
+            Defaults to ``pandas.options.styler.sparse.index`` value.
+        sparse_columns : bool, optional
+            Whether to sparsify the display of a hierarchical index. Setting to False
+            will display each explicit level element in a hierarchical key for each row.
+            Defaults to ``pandas.options.styler.sparse.columns`` value.
+        multirow_align : {"c", "t", "b"}
+            If sparsifying hierarchical MultiIndexes whether to align text centrally,
+            at the top or bottom.
+        multicol_align : {"r", "c", "l"}
+            If sparsifying hierarchical MultiIndex columns whether to align text at
+            the left, centrally, or at the right.
+        siunitx : bool, default False
+            Set to ``True`` to structure LaTeX compatible with the {siunitx} package.
+        encoding : str, default "utf-8"
+            Character encoding setting.
+
+        Returns
+        -------
+        str or None
+            If `buf` is None, returns the result as a string. Otherwise returns `None`.
+
+        See Also
+        --------
+        Styler.format: Format the text display value of cells.
+
+        Notes
+        -----
+        **Latex Packages**
+
+        For the following features we recommend the following LaTeX inclusions:
+
+        ===================== ==========================================================
+        Feature               Inclusion
+        ===================== ==========================================================
+        sparse columns        none: included within default {tabular} environment
+        sparse rows           \\usepackage{multirow}
+        hrules                \\usepackage{booktabs}
+        colors                \\usepackage[table]{xcolor}
+        siunitx               \\usepackage{siunitx}
+        bold (with siunitx)   | \\usepackage{etoolbox}
+                              | \\robustify\\bfseries
+                              | \\sisetup{detect-all = true}  *(within {document})*
+        italic (with siunitx) | \\usepackage{etoolbox}
+                              | \\robustify\\itshape
+                              | \\sisetup{detect-all = true}  *(within {document})*
+        ===================== ==========================================================
+
+        **Cell Styles**
+
+        LaTeX styling can only be rendered if the accompanying styling functions have
+        been constructed with appropriate LaTeX commands. All styling
+        functionality is built around the concept of a CSS ``(<attribute>, <value>)``
+        pair (see `Table Visualization <../../user_guide/style.ipynb>`_), and this
+        should be replaced by a LaTeX
+        ``(<command>, <options>)`` approach. Each cell will be styled individually
+        using nested LaTeX commands with their accompanied options.
+
+        For example the following code will highlight and bold a cell in HTML-CSS:
+
+        >>> df = pd.DataFrame([[1,2], [3,4]])
+        >>> s = df.style.highlight_max(axis=None,
+        ...                            props='background-color:red; font-weight:bold;')
+        >>> s.render()
+
+        The equivalent using LaTeX only commands is the following:
+
+        >>> s = df.style.highlight_max(axis=None,
+        ...                            props='cellcolor:{red}; bfseries: ;')
+        >>> s.to_latex()
+
+        Internally these structured LaTeX ``(<command>, <options>)`` pairs
+        are translated to the
+        ``display_value`` with the default structure:
+        ``\<command><options> <display_value>``.
+        Where there are multiple commands the latter is nested recursively, so that
+        the above example highlighed cell is rendered as
+        ``\cellcolor{red} \bfseries 4``.
+
+        Occasionally this format does not suit the applied command, or
+        combination of LaTeX packages that is in use, so additional flags can be
+        added to the ``<options>``, within the tuple, to result in different
+        positions of required braces (the **default** being the same as ``--nowrap``):
+
+        =================================== ============================================
+        Tuple Format                           Output Structure
+        =================================== ============================================
+        (<command>,<options>)               \\<command><options> <display_value>
+        (<command>,<options> ``--nowrap``)  \\<command><options> <display_value>
+        (<command>,<options> ``--rwrap``)   \\<command><options>{<display_value>}
+        (<command>,<options> ``--wrap``)    {\\<command><options> <display_value>}
+        (<command>,<options> ``--lwrap``)   {\\<command><options>} <display_value>
+        (<command>,<options> ``--dwrap``)   {\\<command><options>}{<display_value>}
+        =================================== ============================================
+
+        For example the `textbf` command for font-weight
+        should always be used with `--rwrap` so ``('textbf', '--rwrap')`` will render a
+        working cell, wrapped with braces, as ``\textbf{<display_value>}``.
+
+        A more comprehensive example is as follows:
+
+        >>> df = pd.DataFrame([[1, 2.2, "dogs"], [3, 4.4, "cats"], [2, 6.6, "cows"]],
+        ...                   index=["ix1", "ix2", "ix3"],
+        ...                   columns=["Integers", "Floats", "Strings"])
+        >>> s = df.style.highlight_max(
+        ...     props='cellcolor:[HTML]{FFFF00}; color:{red};'
+        ...           'textit:--rwrap; textbf:--rwrap;'
+        ... )
+        >>> s.to_latex()
+
+        .. figure:: ../../_static/style/latex_1.png
+
+        **Table Styles**
+
+        Internally Styler uses its ``table_styles`` object to parse the
+        ``column_format``, ``position``, ``position_float``, and ``label``
+        input arguments. These arguments are added to table styles in the format:
+
+        .. code-block:: python
+
+            set_table_styles([
+                {"selector": "column_format", "props": f":{column_format};"},
+                {"selector": "position", "props": f":{position};"},
+                {"selector": "position_float", "props": f":{position_float};"},
+                {"selector": "label", "props": f":{{{label.replace(':','§')}}};"}
+            ], overwrite=False)
+
+        Exception is made for the ``hrules`` argument which, in fact, controls all three
+        commands: ``toprule``, ``bottomrule`` and ``midrule`` simultaneously. Instead of
+        setting ``hrules`` to ``True``, it is also possible to set each
+        individual rule definition, by manually setting the ``table_styles``,
+        for example below we set a regular ``toprule``, set an ``hline`` for
+        ``bottomrule`` and exclude the ``midrule``:
+
+        .. code-block:: python
+
+            set_table_styles([
+                {'selector': 'toprule', 'props': ':toprule;'},
+                {'selector': 'bottomrule', 'props': ':hline;'},
+            ], overwrite=False)
+
+        If other ``commands`` are added to table styles they will be detected, and
+        positioned immediately above the '\\begin{tabular}' command. For example to
+        add odd and even row coloring, from the {colortbl} package, in format
+        ``\rowcolors{1}{pink}{red}``, use:
+
+        .. code-block:: python
+
+            set_table_styles([
+                {'selector': 'rowcolors', 'props': ':{1}{pink}{red};'}
+            ], overwrite=False)
+
+        A more comprehensive example using these arguments is as follows:
+
+        >>> df.columns = pd.MultiIndex.from_tuples([
+        ...     ("Numeric", "Integers"),
+        ...     ("Numeric", "Floats"),
+        ...     ("Non-Numeric", "Strings")
+        ... ])
+        >>> df.index = pd.MultiIndex.from_tuples([
+        ...     ("L0", "ix1"), ("L0", "ix2"), ("L1", "ix3")
+        ... ])
+        >>> s = df.style.highlight_max(
+        ...     props='cellcolor:[HTML]{FFFF00}; color:{red}; itshape:; bfseries:;'
+        ... )
+        >>> s.to_latex(
+        ...     column_format="rrrrr", position="h", position_float="centering",
+        ...     hrules=True, label="table:5", caption="Styled LaTeX Table",
+        ...     multirow_align="t", multicol_align="r"
+        ... )
+
+        .. figure:: ../../_static/style/latex_2.png
+
+        **Formatting**
+
+        To format values :meth:`Styler.format` should be used prior to calling
+        `Styler.to_latex`, as well as other methods such as :meth:`Styler.hide_index`
+        or :meth:`Styler.hide_columns`, for example:
+
+        >>> s.clear()
+        >>> s.table_styles = []
+        >>> s.caption = None
+        >>> s.format({
+        ...    ("Numeric", "Integers"): '\${}',
+        ...    ("Numeric", "Floats"): '{:.3f}',
+        ...    ("Non-Numeric", "Strings"): str.upper
+        ... })
+        >>> s.to_latex()
+        \begin{tabular}{llrrl}
+        {} & {} & \multicolumn{2}{r}{Numeric} & {Non-Numeric} \\
+        {} & {} & {Integers} & {Floats} & {Strings} \\
+        \multirow[c]{2}{*}{L0} & ix1 & \\$1 & 2.200 & DOGS \\
+         & ix2 & \$3 & 4.400 & CATS \\
+        L1 & ix3 & \$2 & 6.600 & COWS \\
+        \end{tabular}
+        """
+        table_selectors = (
+            [style["selector"] for style in self.table_styles]
+            if self.table_styles is not None
+            else []
+        )
+
+        if column_format is not None:
+            # add more recent setting to table_styles
+            self.set_table_styles(
+                [{"selector": "column_format", "props": f":{column_format}"}],
+                overwrite=False,
+            )
+        elif "column_format" in table_selectors:
+            pass  # adopt what has been previously set in table_styles
+        else:
+            # create a default: set float, complex, int cols to 'r' ('S'), index to 'l'
+            _original_columns = self.data.columns
+            self.data.columns = RangeIndex(stop=len(self.data.columns))
+            numeric_cols = self.data._get_numeric_data().columns.to_list()
+            self.data.columns = _original_columns
+            column_format = "" if self.hidden_index else "l" * self.data.index.nlevels
+            for ci, _ in enumerate(self.data.columns):
+                if ci not in self.hidden_columns:
+                    column_format += (
+                        ("r" if not siunitx else "S") if ci in numeric_cols else "l"
+                    )
+            self.set_table_styles(
+                [{"selector": "column_format", "props": f":{column_format}"}],
+                overwrite=False,
+            )
+
+        if position:
+            self.set_table_styles(
+                [{"selector": "position", "props": f":{position}"}],
+                overwrite=False,
+            )
+
+        if position_float:
+            if position_float not in ["raggedright", "raggedleft", "centering"]:
+                raise ValueError(
+                    f"`position_float` should be one of "
+                    f"'raggedright', 'raggedleft', 'centering', "
+                    f"got: '{position_float}'"
+                )
+            self.set_table_styles(
+                [{"selector": "position_float", "props": f":{position_float}"}],
+                overwrite=False,
+            )
+
+        if hrules:
+            self.set_table_styles(
+                [
+                    {"selector": "toprule", "props": ":toprule"},
+                    {"selector": "midrule", "props": ":midrule"},
+                    {"selector": "bottomrule", "props": ":bottomrule"},
+                ],
+                overwrite=False,
+            )
+
+        if label:
+            self.set_table_styles(
+                [{"selector": "label", "props": f":{{{label.replace(':', '§')}}}"}],
+                overwrite=False,
+            )
+
+        if caption:
+            self.set_caption(caption)
+
+        if sparse_index is None:
+            sparse_index = get_option("styler.sparse.index")
+        if sparse_columns is None:
+            sparse_columns = get_option("styler.sparse.columns")
+
+        latex = self._render_latex(
+            sparse_index=sparse_index,
+            sparse_columns=sparse_columns,
+            multirow_align=multirow_align,
+            multicol_align=multicol_align,
+        )
+
+        return save_to_buffer(latex, buf=buf, encoding=encoding)
+
     def set_td_classes(self, classes: DataFrame) -> Styler:
         """
         Set the DataFrame of strings added to the ``class`` attribute of ``<td>``