Skip to content

Python tool for converting files and office documents to Markdown.

fix: prefer data-src over placeholder data URI in img src (#2417)

Machen Johnb59e0643dc86
diff --git a/packages/markitdown/src/markitdown/converters/_markdownify.py b/packages/markitdown/src/markitdown/converters/_markdownify.py
index 4fff2e2c..e0bbcaf2 100644
--- a/packages/markitdown/src/markitdown/converters/_markdownify.py
+++ b/packages/markitdown/src/markitdown/converters/_markdownify.py
@@ -115,7 +115,18 @@ class _CustomMarkdownify(markdownify.MarkdownConverter):
"""Same as usual converter, but removes data URIs"""
alt = el.attrs.get("alt", None) or ""
- src = el.attrs.get("src", None) or el.attrs.get("data-src", None) or ""
+ src = el.attrs.get("src", None) or ""
+ data_src = el.attrs.get("data-src", None) or ""
+ # Lazy-loading libraries commonly leave a tiny placeholder data URI in
+ # src and put the real image in data-src. Prefer data-src when src
+ # isn't a usable URL, so the placeholder doesn't win over actual
+ # content. When keep_data_uris is set the caller explicitly wants the
+ # embedded bytes, so a data URI in src is left alone.
+ if data_src and (
+ not src
+ or (src[:5].lower() == "data:" and not self.options["keep_data_uris"])
+ ):
+ src = data_src
title = el.attrs.get("title", None) or ""
title_part = ' "%s"' % title.replace('"', r"\"") if title else ""
# Remove all line breaks from alt
diff --git a/packages/markitdown/tests/test_html_converter.py b/packages/markitdown/tests/test_html_converter.py
index e3e62b51..e5643519 100644
--- a/packages/markitdown/tests/test_html_converter.py
+++ b/packages/markitdown/tests/test_html_converter.py
@@ -3,10 +3,11 @@ import io
from markitdown import MarkItDown
-def _convert_html(html: str) -> str:
+def _convert_html(html: str, **kwargs) -> str:
result = MarkItDown().convert_stream(
io.BytesIO(html.encode("utf-8")),
file_extension=".html",
+ **kwargs,
)
return result.markdown
@@ -63,3 +64,61 @@ def test_html_href_does_not_quote_query_or_fragment() -> None:
markdown = _convert_html(f'<a href="{href}">example</a>')
assert f"[example]({expected_href})" in markdown
+
+
+def test_img_prefers_data_src_over_placeholder_data_uri() -> None:
+ placeholder = (
+ "data:image/gif;base64,R0lGODlhAQABAIAAAAAAAP///yH5BAEAAAAALAAAAAABAAEAAAIBTAA7"
+ )
+ real_src = "https://example.com/photo.jpg"
+ html = (
+ f'<img src="{placeholder}" data-src="{real_src}" alt="A photo" loading="lazy">'
+ )
+
+ markdown = _convert_html(html)
+
+ assert f"![A photo]({real_src})" in markdown
+ assert placeholder not in markdown
+
+
+def test_img_uses_real_src_over_data_src_when_both_present() -> None:
+ real_src = "https://example.com/photo.jpg"
+ other_src = "https://example.com/photo-alt.jpg"
+ html = f'<img src="{real_src}" data-src="{other_src}" alt="A photo">'
+
+ markdown = _convert_html(html)
+
+ assert f"![A photo]({real_src})" in markdown
+
+
+def test_img_falls_back_to_data_src_when_src_missing() -> None:
+ real_src = "https://example.com/photo.jpg"
+ html = f'<img data-src="{real_src}" alt="A photo">'
+
+ markdown = _convert_html(html)
+
+ assert f"![A photo]({real_src})" in markdown
+
+
+def test_img_keeps_truncated_data_uri_when_no_data_src() -> None:
+ placeholder = (
+ "data:image/gif;base64,R0lGODlhAQABAIAAAAAAAP///yH5BAEAAAAALAAAAAABAAEAAAIBTAA7"
+ )
+ html = f'<img src="{placeholder}" alt="A photo">'
+
+ markdown = _convert_html(html)
+
+ assert "![A photo](data:image/gif;base64...)" in markdown
+
+
+def test_img_keeps_embedded_data_uri_over_data_src_when_keeping_data_uris() -> None:
+ embedded = (
+ "data:image/gif;base64,R0lGODlhAQABAIAAAAAAAP///yH5BAEAAAAALAAAAAABAAEAAAIBTAA7"
+ )
+ other_src = "https://example.com/photo.jpg"
+ html = f'<img src="{embedded}" data-src="{other_src}" alt="A photo">'
+
+ markdown = _convert_html(html, keep_data_uris=True)
+
+ assert f"![A photo]({embedded})" in markdown
+ assert other_src not in markdown