SVGObject.from_file: use the encoding the file declares

CHANGELOG.md

@@ -23,6 +23,7 @@ This can also be enabled programmatically with `warnings.simplefilter('default',
 ### Fixed
 * SVG: `currentColor` and `inherit` are now read regardless of case, like the color keywords, so `fill="currentcolor"`, as in the Font Awesome icons of many web pages, no longer fails with `ValueError: unsupported color specification currentcolor` - thanks to @RavSinghChandan
 * SVG: black or gray text no longer fails with an `AssertionError` when it is in a `<tspan>` that has its own style, as in every text written by Inkscape: the check on the text color left out `DeviceGray`, the color fpdf2 uses for black and gray - thanks to @RavSinghChandan
+* `SVGObject.from_file()` reads a file in the encoding it declares, such as `<?xml version="1.0" encoding="iso-8859-1"?>`, like `FPDF.image()` does, instead of always decoding it as UTF-8, which failed with a `UnicodeDecodeError`; an `encoding` given explicitly is still used - thanks to @RavSinghChandan
 * SVG: an image without an `xmlns` declaration no longer fails with `ValueError: substring not found` when it contains a tag that is not supported, such as `<title>` or `<script>`; the tag is skipped with a warning, as in an image that declares the namespace - thanks to @RavSinghChandan
 * `FPDF.text()` raises a clear `ValueError` for a NaN or infinite `x` or `y`, like `FPDF.set_xy()`, instead of writing `nan` or `inf` into the page, which is not a valid PDF number - thanks to @RavSinghChandan
 * SVG: `clip-path="none"`, as written by Inkscape, no longer fails with an `AssertionError`, and a `url(#id)` reference to a clip path whose id contains a `-` or a `.`, or that is quoted, now finds that clip path, instead of failing with a `KeyError` or clipping with another one whose id is the part before the `-` or `.` - thanks to @RavSinghChandan

fpdf/svg.py

@@ -1113,7 +1113,7 @@ def from_file(
         cls,
         filename: str | PathLike[str],
         *args: Any,
-        encoding: str = "utf-8",
+        encoding: Optional[str] = None,
         **kwargs: Any,
     ) -> "SVGObject":
         """
@@ -1123,11 +1123,17 @@ def from_file(
             filename (path-like): the path to a file containing SVG data.
             *args: forwarded directly to the SVGObject initializer. For subclass use.
             encoding (str): optional charset encoding to use when reading the file.
+                By default, the encoding declared by the file is used, as in
+                `FPDF.image()`, and UTF-8 if it declares none.
             **kwargs: forwarded directly to the SVGObject initializer. For subclass use.
 
         Returns:
             A converted `SVGObject`.
         """
+        if encoding is None:
+            # The XML parser reads the encoding declaration of the file
+            with open(filename, "rb") as svgfile:
+                return cls(svgfile.read(), *args, **kwargs)
         with open(filename, "r", encoding=encoding) as svgfile:
             return cls(svgfile.read(), *args, **kwargs)
 

test/svg/test_svg.py

@@ -546,6 +546,32 @@ def test_svg_text_gray_color_in_styled_tspan(
         svg.draw_to_page(pdf)
         assert f"{color_operators} [] 0 d BT" in bytes(pdf.output()).decode("latin-1")
 
+    @pytest.mark.parametrize(
+        "declaration, file_encoding, from_file_kwargs",
+        [
+            ('<?xml version="1.0" encoding="iso-8859-1"?>', "latin-1", {}),
+            ("", "utf-8", {}),
+            ("", "latin-1", {"encoding": "latin-1"}),
+        ],
+    )
+    def test_svg_from_file_encoding(
+        self, tmp_path, declaration, file_encoding, from_file_kwargs
+    ):
+        svg_path = tmp_path / "text.svg"
+        svg_path.write_bytes(
+            (
+                f'{declaration}<svg xmlns="http://www.w3.org/2000/svg" width="50" '
+                'height="20"><text x="1" y="10" font-family="helvetica">Z\u00fcrich'
+                "</text></svg>"
+            ).encode(file_encoding)
+        )
+        svg = fpdf.svg.SVGObject.from_file(svg_path, **from_file_kwargs)
+        pdf = fpdf.FPDF()
+        pdf.compress = False
+        pdf.add_page()
+        svg.draw_to_page(pdf)
+        assert "(Z\u00fcrich) Tj" in bytes(pdf.output()).decode("latin-1")
+
     def test_svg_symbol(self):
         svg_data = (
             '<svg xmlns="http://www.w3.org/2000/svg" '