Add basic support for adding SVG pictures to docx files #1343 by takis · Pull Request #1386

Add basic support for adding SVG pictures to docx files #1343 by takis · Pull Request #1386 · python-openxml/python-docx

takis mentioned this pull request

May 2, 2024

Hello, Is there a target date for this code to be committed to the master ?

I really need this implementation.

Thanks for the awesome software.

Hi @marcuoli,

No, unfortunately not.

I do try to keep my fork in sync with the main repository.

I hope this patch will be merged soon.

I'm trying to use your patch, but it doesn't appear to work with matplotlib figures since their sizes are in pt rather than pixels.

Example:

import docx
from docx.shared import Inches

import numpy as np
import matplotlib.pyplot as plt
from io import BytesIO

from tempfile import mktemp

x = np.linspace(-2*np.pi, 2*np.pi, 100)

fig, ax = plt.subplots()
ax.plot(x, np.sin(x), color="C0")
ax.plot(x, np.cos(x), color="C1")

buf = BytesIO()
fig.savefig(buf, format="svg")
w, h = (Inches(s) for s in fig.get_size_inches())

p = mktemp(suffix=".svg")
fig.savefig(p)

# doesn't work
doc = docx.Document()
doc.add_picture(buf, w, h)

Traceback:

File c:\Users\rpr\.conda\envs\funandgames\Lib\site-packages\docx\image\svg.py:43, in Svg._dimensions_from_stream(cls, stream)
     40 root = ET.fromstring(data)
     41 # FIXME: The width could be expressed as '4cm'
     42 # See https://www.w3.org/TR/SVG11/struct.html#NewDocument
---> 43 width = int(root.attrib["width"])
     44 height = int(root.attrib["height"])
     45 return width, height

I think a pt is 1.33x a pixel in svgs. If so, it might be easy to at least handle pt. But maybe it depends on dpi?
For matplotlib at least, I think dpi is fixed at 72 (e.g. here)

An example of an mpl svg is attached.

With this patch I can insert mpl figures. It's written rather verbose, as I don't really know the style of python-docx.
@takis would you consider adding this patch or something similar?

From 458df2e57c57be410e6d2486ae67e44cebbd177e Mon Sep 17 00:00:00 2001
From: Rasmus <git@pank.eu>
Date: Fri, 18 Oct 2024 10:02:49 +0200
Subject: [PATCH] svg: Handle non-pixels units

---
 src/docx/image/svg.py | 32 ++++++++++++++++++++++++++++----
 1 file changed, 28 insertions(+), 4 deletions(-)

diff --git a/src/docx/image/svg.py b/src/docx/image/svg.py
index 0247a77..4be43e1 100644
--- a/src/docx/image/svg.py
+++ b/src/docx/image/svg.py
@@ -33,13 +33,37 @@ class Svg(BaseImageHeader):
         """
         return "svg"
 
+    @classmethod
+    def _dimension_to_px(cls, s):
+        """Convert SVG unit *s* to pixels.
+
+        Based on the `table`_ on absolute length in W3 CSS Values 3.
+
+        .. table: https://www.w3.org/TR/css-values-3/#absolute-lengths
+        """
+        s = str(s)
+        unit = s.lstrip("0123456789.").casefold()
+        snumber = s[:-len(unit)] if len(unit) > 0 else s
+        number = float(snumber)
+        inch = 96
+        cm = inch/2.54
+        units = {"cm": cm,
+                 "mm": cm/10,
+                 "q": cm/40,  # 'Q' in W3 table
+                 "in": inch,
+                 "pc": inch/6,
+                 "pt": inch/72,
+                 "px": 1}
+        # Unitless is pixels
+        number_px = number * units.get(unit, 1)
+        return int(number_px)
+
     @classmethod
     def _dimensions_from_stream(cls, stream):
         stream.seek(0)
         data = stream.read()
         root = ET.fromstring(data)
-        # FIXME: The width could be expressed as '4cm'
-        # See https://www.w3.org/TR/SVG11/struct.html#NewDocument
-        width = int(root.attrib["width"])
-        height = int(root.attrib["height"])
+        # The width could be expressed as '4cm'
+        width = cls._dimension_to_px(root.attrib["width"])
+        height = cls._dimension_to_px(root.attrib["height"])
         return width, height
-- 
2.44.0.windows.1
```

lsaint left a comment •

Loading

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

This PR broke a unitest which need to add xmlns:asvg="http://schemas.microsoft.com/office/drawing/2016/SVG/main
to
tests/test_files/snippets/inline.txt.

some of svg files without "width" and "height" not supported yet.
eg:
<svg viewBox="0 0 96 96" xmlns="http://www.w3.org/2000/svg" xmlns:xlink="http://www.w3.org/1999/xlink" id="Icons_CheckboxCrossed" overflow="hidden"><path d="M37.076 62.948 48.008 52.201 58.939 62.948 63.146 58.67 52.286 47.994 63.146 37.318 58.939 33.04 48.008 43.787 37.076 33.04 32.87 37.318 43.729 47.994 32.87 58.67 37.076 62.948Z"/><path d="M21 21 21 75 75 75 75 21ZM69 69 27 69 27 27 69 27Z"/></svg>

I have fixed them and added some unittests in my private fork

@sgtyfianalisisfinanciero
Could you share the MPL code of something that doesn't work? Then I will be happy to see if I can reproduce it.

AI've written an package that hotpatches python-docx to support svgs based on this thread that I use for work until svg support is merged. Would that be interesting for others?

@pank nervermind, the problem was not with the images or the mpl code that generated it. Trying to elaborate a minimal reproducible example I catched the issue

My workflow was something like that:

Paste images in some table cells (works)
Generate some other (totally independent) tables, saving the partial document after each table generation for debugging purposes
Paste some other images in some table cells (placeholder shown)

Removing the saving step solves the issue. I'm clueless about why, but it's fine

On the other hand I think your hotfix package would be very interesting for a lot of people. While this get merged, everyone who needed that feature could just uv add git+... your fork, which is nicer than hardcoding your patch (which can easily get hard to find in the conversation) into the .venv folder.

This is a really important feature, what are the steps needed prior to a merge? @takis, @lsaint, @scanny

@sgtyfianalisisfinanciero sorry about the delay. I wanted to add tests and read through the code, but didn't manage before my going on vacation. I still haven't managed, but have added the hotpatch package here:

https://github.com/pank/python-docx-svg

I haven't added it to PyPi, but could do so if that would be useful. It can still be installed via pip.

And more news on this? It seems like the patch has addressed the issue, but is still waiting to be merged?

lsaint left a comment • Loading

Choose a reason for hiding this comment

lsaint left a comment •

Loading