Jules 1.23.20

.github/ISSUE_TEMPLATE/bug_report.yml

@@ -47,6 +47,7 @@ body:
       label: PyMuPDF version
       options:
         - 
+        - 1.23.20
         - 1.23.19
         - 1.23.18
         - 1.23.17

changes.txt

@@ -2,6 +2,17 @@ Change Log
 ==========
 
 
+**Changes in version 1.23.20 (2024-01-29)**
+
+* Bug fixes:
+
+  * **Fixed** `3100 <https://github.com/pymupdf/PyMuPDF/issues/3100>`_: Wrong internal property accessed in get_xml_metadata
+
+* Other:
+
+  * Significantly improved speed of `Document.get_toc()`.
+
+
 **Changes in version 1.23.19 (2024-01-25)**
 
 * Bug fixes:

docs/version.rst

@@ -1,6 +1,6 @@
 ----
 
-This documentation covers **PyMuPDF v1.23.19** features as of **2024-01-25 00:00:01**.
+This documentation covers **PyMuPDF v1.23.20** features as of **2024-01-29 00:00:01**.
 
 The major and minor versions of **PyMuPDF** and **MuPDF** will always be the same. Only the third qualifier (patch level) may deviate from that of **MuPDF**.
 

setup.py

@@ -1072,7 +1072,7 @@ def sdist():
 # We generate different wheels depending on g_flavour.
 #
 
-version = '1.23.19'
+version = '1.23.20'
 version_b = '1.23.9'
 
 tag_python = None

src/__init__.py

@@ -4271,7 +4271,7 @@ def get_xml_metadata(self):
                     PDF_NAME('Root'),
                     PDF_NAME('Metadata'),
                     )
-        if xml and xml.internal:
+        if xml and xml.m_internal:
             buff = mupdf.pdf_load_stream(xml)
             rc = JM_UnicodeFromBuffer(buff)
         else:
@@ -7808,36 +7808,50 @@ def _insert_image(self,
                     do_have_imask = 0
 
         if do_have_imask:
-            #log( 'do_have_imask')
-            # mupdf.FzCompressedBuffer is not copyable, so
-            # mupdf.fz_compressed_image_buffer() does not work - it cannot
-            # return by value. And sharing a fz_compressed_buffer betwen two
-            # `fz_image`'s doesn't work, so we use a raw fz_compressed_buffer
-            # here, not a mupdf.FzCompressedBuffer.
-            #
-            cbuf1 = mupdf.ll_fz_compressed_image_buffer( image.m_internal)
-            if not cbuf1:
-                raise ValueError( "uncompressed image cannot have mask")
-            bpc = image.bpc()
-            colorspace = image.colorspace()
-            xres, yres = mupdf.fz_image_resolution(image)
-            mask = mupdf.fz_new_image_from_buffer(maskbuf)
             if mupdf_version_tuple >= (1, 24):
-                zimg = mupdf.ll_fz_new_image_from_compressed_buffer2(
+                # `fz_compressed_buffer` is reference counted and
+                # `mupdf.fz_new_image_from_compressed_buffer2()`
+                # is povided as a Swig-friendly wrapper for
+                # `fz_new_image_from_compressed_buffer()`, so we can do things
+                # straightfowardly.
+                #
+                cbuf1 = mupdf.fz_compressed_image_buffer( image)
+                if not cbuf1.m_internal:
+                    raise ValueError( "uncompressed image cannot have mask")
+                bpc = image.bpc()
+                colorspace = image.colorspace()
+                xres, yres = mupdf.fz_image_resolution(image)
+                mask = mupdf.fz_new_image_from_buffer(maskbuf)
+                image = mupdf.fz_new_image_from_compressed_buffer2(
                         w,
                         h,
                         bpc,
-                        colorspace.m_internal,
+                        colorspace,
                         xres,
                         yres,
                         1,  # interpolate
                         0,  # imagemask,
                         None,   # decode
                         None,   # colorkey
                         cbuf1,
-                        mask.m_internal,
+                        mask,
                         )
             else:
+                #log( 'do_have_imask')
+                # mupdf.FzCompressedBuffer is not copyable, so
+                # mupdf.fz_compressed_image_buffer() does not work - it cannot
+                # return by value. And sharing a fz_compressed_buffer betwen two
+                # `fz_image`'s doesn't work, so we use a raw fz_compressed_buffer
+                # here, not a mupdf.FzCompressedBuffer.
+                #
+                cbuf1 = mupdf.ll_fz_compressed_image_buffer( image.m_internal)
+                if not cbuf1:
+                    raise ValueError( "uncompressed image cannot have mask")
+                bpc = image.bpc()
+                colorspace = image.colorspace()
+                xres, yres = mupdf.fz_image_resolution(image)
+                mask = mupdf.fz_new_image_from_buffer(maskbuf)
+
                 # mupdf.ll_fz_new_image_from_compressed_buffer() is not usable.
                 zimg = extra.fz_new_image_from_compressed_buffer(
                         w,
@@ -7852,26 +7866,26 @@ def _insert_image(self,
                         mask.m_internal,
                         )
 
-            zimg = mupdf.FzImage(zimg)
-
-            # `image` and `zimage` both have pointers to the same
-            # `fz_compressed_buffer`, which is not reference counted, and they
-            # both think that they own it.
-            #
-            # So we do what the classic implementataion does, and simply ensure
-            # that `fz_drop_image(image)` is never called. This will leak
-            # some of `image`'s allocations (for example the main `fz_image`
-            # allocation), but it's not trivial to avoid this.
-            #
-            # Perhaps we could manually set `fz_image`'s
-            # `fz_compressed_buffer*` to null? Trouble is we'd have to
-            # cast the `fz_image*` to a `fz_compressed_image*` to see the
-            # `fz_compressed_buffer*`, which is probably not possible from
-            # Python?
-            #
-            image.m_internal = None
+                zimg = mupdf.FzImage(zimg)
+
+                # `image` and `zimage` both have pointers to the same
+                # `fz_compressed_buffer`, which is not reference counted, and they
+                # both think that they own it.
+                #
+                # So we do what the classic implementataion does, and simply ensure
+                # that `fz_drop_image(image)` is never called. This will leak
+                # some of `image`'s allocations (for example the main `fz_image`
+                # allocation), but it's not trivial to avoid this.
+                #
+                # Perhaps we could manually set `fz_image`'s
+                # `fz_compressed_buffer*` to null? Trouble is we'd have to
+                # cast the `fz_image*` to a `fz_compressed_image*` to see the
+                # `fz_compressed_buffer*`, which is probably not possible from
+                # Python?
+                #
+                image.m_internal = None
 
-            image = zimg
+                image = zimg
 
         if do_have_image:
             #log( 'do_have_image')
@@ -21801,8 +21815,8 @@ def int_rc(text):
     return int(text)
 
 VersionFitz = "1.23.9" # MuPDF version.
-VersionBind = "1.23.19" # PyMuPDF version.
-VersionDate = "2024-01-25 00:00:01"
+VersionBind = "1.23.20" # PyMuPDF version.
+VersionDate = "2024-01-29 00:00:01"
 VersionDate2 = VersionDate.replace('-', '').replace(' ', '').replace(':', '')
 version = (VersionBind, VersionFitz, VersionDate2)
 pymupdf_version_tuple = tuple( [int_rc(i) for i in VersionBind.split('.')])

src_classic/version.i

@@ -1,7 +1,7 @@
 %pythoncode %{
 VersionFitz = "1.23.9" # MuPDF version.
-VersionBind = "1.23.19" # PyMuPDF version.
-VersionDate = "2024-01-25 00:00:01"
-version = (VersionBind, VersionFitz, "20240125000001")
+VersionBind = "1.23.20" # PyMuPDF version.
+VersionDate = "2024-01-29 00:00:01"
+version = (VersionBind, VersionFitz, "20240129000001")
 pymupdf_version_tuple = tuple( [int(i) for i in VersionFitz.split('.')])
 %}

tests/test_general.py

@@ -863,3 +863,8 @@ def next_fd():
     assert fd2 == fd1 + 1
     assert fd3 == fd1
     assert fd4 == fd1
+
+def test_xml():
+    path = os.path.abspath(f'{__file__}/../../tests/resources/2.pdf')
+    with fitz.open(path) as document:
+        document.get_xml_metadata()