Using .local_context() in .write_html() to better isolate HTML render…

…ing & styling

CHANGELOG.md

@@ -28,6 +28,7 @@ This can also be enabled programmatically with `warnings.simplefilter('default',
 ### Fixed
 * [`FPDF.local_context()`](https://py-pdf.github.io/fpdf2/fpdf/fpdf.html#fpdf.fpdf.FPDF.local_context) used to leak styling during page breaks, when rendering `footer()` & `header()`
 * [`fpdf.drawing.DeviceCMYK`](https://py-pdf.github.io/fpdf2/fpdf/drawing.html#fpdf.drawing.DeviceCMYK) objects can now be passed to [`FPDF.set_draw_color()`](https://py-pdf.github.io/fpdf2/fpdf/fpdf.html#fpdf.fpdf.FPDF.set_draw_color), [`FPDF.set_fill_color()`](https://py-pdf.github.io/fpdf2/fpdf/fpdf.html#fpdf.fpdf.FPDF.set_fill_color) and [`FPDF.set_text_color()`](https://py-pdf.github.io/fpdf2/fpdf/fpdf.html#fpdf.fpdf.FPDF.set_text_color) without raising a `ValueError`: [documentation](https://py-pdf.github.io/fpdf2/Text.html#text-formatting).
+* [`FPDF.write_html()`](https://py-pdf.github.io/fpdf2/fpdf/fpdf.html#fpdf.fpdf.FPDF.write_html): fixing rendering of `<hr>` tags, that do not trigger a page break anymore
 * individual `/Resources` directories are now properly created for each document page. This change ensures better compliance with the PDF specification but results in a slight increase in the size of PDF documents. You can still use the old behavior by setting `FPDF().single_resources_object = True`
 * line size calculation for fragments when text shaping is used
 * fixed incoherent indentation of long list entries - _cf._ [issue #1073](https://github.com/py-pdf/fpdf2/issues/1073)

fpdf/fpdf.py

@@ -434,7 +434,8 @@ def write_html(self, text, *args, **kwargs):
             tag_styles (dict): mapping of HTML tag names to colors
         """
         html2pdf = self.HTML2FPDF_CLASS(self, *args, **kwargs)
-        html2pdf.feed(text)
+        with self.local_context():
+            html2pdf.feed(text)
 
     def _set_min_pdf_version(self, version):
         self.pdf_version = max(self.pdf_version, version)
@@ -2797,12 +2798,12 @@ def local_context(self, **kwargs):
                 "cannot create a local context inside an unbreakable() code block"
             )
         self._push_local_stack()
-        self._init_local_context(**kwargs)  # write "q" in the output stream
+        self._start_local_context(**kwargs)
         yield
-        self._out("Q")
+        self._end_local_context()
         self._pop_local_stack()
 
-    def _init_local_context(
+    def _start_local_context(
         self,
         font_family=None,
         font_style=None,
@@ -2815,7 +2816,7 @@ def _init_local_context(
         **kwargs,
     ):
         """
-        This method starts a "q" context in the output stream,
+        This method starts a "q/Q" context in the page content stream,
         and inserts operators in it to initialize all the PDF settings specified.
         """
         if "font_size_pt" in kwargs:
@@ -2884,6 +2885,12 @@ def _init_local_context(
         if dash_pattern is not None:
             self.set_dash_pattern(**dash_pattern)
 
+    def _end_local_context(self):
+        """
+        This method ends a "q/Q" context in the page content stream.
+        """
+        self._out("Q")
+
     @property
     def accept_page_break(self):
         """
@@ -3575,14 +3582,19 @@ def _perform_page_break_if_need_be(self, h):
 
     def _perform_page_break(self):
         x = self.x
+        # If we are in a .local_context(), we need to temporarily leave it,
+        # by popping out every GraphicsState:
         gs_stack = []
         while self._is_current_graphics_state_nested():
-            self._out("Q")
+            # This code assumes that every Graphics State in the stack
+            # has been pushed in it while adding a "q" in the PDF stream
+            # (which is what FPDF.local_context() does):
+            self._end_local_context()
             gs_stack.append(self._pop_local_stack())
         self.add_page(same=True)
         for prev_gs in reversed(gs_stack):
             self._push_local_stack(prev_gs)
-            self._init_local_context(**prev_gs)
+            self._start_local_context(**prev_gs)
         self.x = x  # restore x but not y after drawing header
 
     def _has_next_page(self):
@@ -4994,7 +5006,7 @@ def insert_toc_placeholder(self, render_toc_function, pages=1):
             render_toc_function, self.page, self.y, pages
         )
         for _ in range(pages):
-            self.add_page()
+            self._perform_page_break()
 
     def set_section_title_styles(
         self,

fpdf/html.py

@@ -325,13 +325,10 @@ def __init__(
         self.emphasis = dict(b=False, i=False, u=False)
         self.font_size = pdf.font_size_pt
         self.set_font(pdf.font_family or "times", size=self.font_size, set_default=True)
-        self._prev_font = (pdf.font_family, self.font_size, self.emphasis)
-        self.pdf._push_local_stack()  # xpylint: disable=protected-access
 
         self._pre_formatted = False  # preserve whitespace while True.
-        self._pre_started = (
-            False  # nothing written yet to <pre>, remove one initial nl.
-        )
+        # nothing written yet to <pre>, remove one initial nl:
+        self._pre_started = False
         self.follows_trailing_space = False  # The last write has ended with a space.
         self.follows_heading = False  # We don't want extra space below a heading.
         self.href = ""
@@ -464,11 +461,7 @@ def _end_paragraph(self):
         self.align = ""
         if self._paragraph:
             self._column.end_paragraph()
-            our_context = (
-                self.pdf._pop_local_stack()  # pylint: disable=protected-access
-            )
             self._column.render()
-            self.pdf._push_local_stack(our_context)  # pylint: disable=protected-access
             self._paragraph = None
             self.follows_trailing_space = True
 
@@ -530,15 +523,12 @@ def handle_data(self, data):
         elif self._pre_formatted:  # pre blocks
             # If we want to mimick the exact HTML semantics about newlines at the
             # beginning and end of the block, then this needs some more thought.
-            s_nl = data.startswith("\n") and self._pre_started
+            if data.startswith("\n") and self._pre_started:
+                if data.endswith("\n"):
+                    data = data[1:-1]
+                else:
+                    data = data[1:]
             self._pre_started = False
-            e_nl = data.endswith("\n")
-            if s_nl and e_nl:
-                data = data[1:-1]
-            elif s_nl:
-                data = data[1:]
-            # elif e_nl:
-            #    data = data[:-1]
             self._write_data(data)
         else:
             data = _WS_SUB_PAT.sub(" ", data)
@@ -650,7 +640,13 @@ def handle_starttag(self, tag, attrs):
                 size=tag_style.size_pt or self.font_size,
             )
         if tag == "hr":
-            self.pdf.add_page(same=True)
+            self.pdf.line(
+                x1=self.pdf.l_margin,
+                y1=self.pdf.y,
+                x2=self.pdf.l_margin + self.pdf.epw,
+                y2=self.pdf.y,
+            )
+            self._write_paragraph("\n")
         if tag == "code":
             self.style_stack.append(
                 FontFace(
@@ -667,6 +663,7 @@ def handle_starttag(self, tag, attrs):
                 size=tag_style.size_pt or self.font_size,
             )
         if tag == "pre":
+            self._end_paragraph()
             self.style_stack.append(
                 FontFace(
                     family=self.font_family,
@@ -682,8 +679,8 @@ def handle_starttag(self, tag, attrs):
                 size=tag_style.size_pt or self.font_size,
             )
             self._pre_formatted = True
-            self._new_paragraph()
             self._pre_started = True
+            self._new_paragraph()
         if tag == "blockquote":
             tag_style = self.tag_styles[tag]
             if tag_style.color:
@@ -928,12 +925,12 @@ def handle_endtag(self, tag):
             self.set_font(font_face.family, font_face.size_pt)
             self.set_text_color(*font_face.color.colors255)
         if tag == "pre":
-            self._end_paragraph()
             font_face = self.style_stack.pop()
             self.set_font(font_face.family, font_face.size_pt)
             self.set_text_color(*font_face.color.colors255)
             self._pre_formatted = False
             self._pre_started = False
+            self._end_paragraph()
         if tag == "blockquote":
             self._end_paragraph()
             self.set_text_color(*self.font_color)
@@ -991,10 +988,6 @@ def feed(self, data):
         while self._tags_stack and self._tags_stack[-1] in self.HTML_UNCLOSED_TAGS:
             self._tags_stack.pop()
         self._end_paragraph()  # render the final chunk of text and clean up our local context.
-        self.pdf._pop_local_stack()  # pylint: disable=protected-access
-        if self._prev_font[0]:  # restore previously defined font settings
-            self.emphasis = self._prev_font[2]
-            self.set_font(self._prev_font[0], size=self._prev_font[1], set_default=True)
         if self._tags_stack and self.warn_on_tags_not_matching:
             LOGGER.warning("Missing HTML end tag for <%s>", self._tags_stack[-1])
 

fpdf/text_region.py

@@ -718,7 +718,7 @@ def render(self):
         _first_page_top = max(self.pdf.t_margin, self.pdf.y)
         self._render_page_lines(text_lines, _first_page_top, page_bottom)
         while text_lines:
-            self.pdf.add_page(same=True)
+            self.pdf._perform_page_break()
             self._cur_column = 0
             self._render_page_lines(text_lines, self.pdf.y, page_bottom)
 

scripts/compare-changed-pdfs.py

@@ -3,9 +3,9 @@
 # Generate a HTML page that makes it easy to visually compare all PDF files
 # that are modified in the current branch, compared to the master branch.
 
-# USAGE: ./compare-changed-pdfs.py
+# USAGE: ./compare-changed-pdfs.py [test_subdir_path]
 
-import webbrowser
+import sys, webbrowser
 from functools import partial
 from http.server import HTTPServer, SimpleHTTPRequestHandler
 from os import makedirs, scandir
@@ -30,15 +30,18 @@ def scantree_dirs(path):
             yield from scantree_dirs(entry.path)
 
 
+target_dir = sys.argv[1] if len(sys.argv) > 1 else "test"
+print(f"Processing all PDF reference files in {target_dir}")
+
 stdout = check_output("git diff --name-status master", shell=True)
 changed_pdf_files = [
     line[1:].strip()
     for line in stdout.decode("utf-8").splitlines()
-    if line.startswith("M\ttest/")
+    if line.startswith(f"M\t{target_dir}")
 ]
 
 TMP_DIR.mkdir(exist_ok=True)
-for dir in scantree_dirs(REPO_DIR / "test"):
+for dir in scantree_dirs(REPO_DIR / target_dir):
     (TMP_DIR / dir).mkdir(exist_ok=True)
 for changed_pdf_file in changed_pdf_files:
     command = f"git show master:{changed_pdf_file} > {TMP_DIR}/{changed_pdf_file}"

test/html/test_html.py

@@ -53,10 +53,11 @@ def test_html_features(tmp_path):
     pdf.write_html("<h4>h4</h4>")
     pdf.write_html("<h5>h5</h5>")
     pdf.write_html("<h6>h6</h6>")
-    pdf.write_html("<br />")
-    pdf.write_html("<hr />")
-    pdf.write_html("<br />")
-    pdf.write_html("<br />")
+    pdf.write_html("<p>Rendering &lt;hr&gt;:</p>")
+    pdf.write_html("<hr>")
+    # Now inserting <br> tags until a page jump is triggered:
+    for _ in range(25):
+        pdf.write_html("<br>")
     pdf.write_html("<pre>i am preformatted text.</pre>")
     pdf.write_html("<blockquote>hello blockquote</blockquote>")
     pdf.write_html("<ul><li>li1</li><li>another</li><li>l item</li></ul>")
@@ -427,10 +428,11 @@ def test_html_whitespace_handling(tmp_path):  # Issue 547
 <p>
 <b>Testing   </b> paragraph blocks
         that <i>span</i> <b>multiple lines</b>.
-    Testing tabs       and    spaces<br>
-    and break tags.<br>
+    Testing tabs       and    spaces
+    <br>and break tags.
 </p>
-<code>Testing code blocks with tabs      and    spaces.</code><br>
+<code>Testing code blocks with tabs      and    spaces.</code>
+<br>
 <pre>
 Testing pre blocks
 that span multiple lines
@@ -444,8 +446,8 @@ def test_html_whitespace_handling(tmp_path):  # Issue 547
 </code></pre>
 
 <p>Testing unicode nbsp \u00a0\u00a0\u00a0\u00a0,
-and html nbsp &nbsp;&nbsp;&nbsp;&nbsp;.<br>
-    \u00a0&nbsp;&nbsp;Testing leading nbsp
+and html nbsp &nbsp;&nbsp;&nbsp;&nbsp;.
+<br>\u00a0&nbsp;&nbsp;Testing leading nbsp
 </p>
 </body>
 """

test/outline/test_outline_html.py

@@ -251,7 +251,7 @@ def render_toc(self, pdf, outline):
             )
 
 
-def test_custom_HTML2FPDF(tmp_path):  # issue 240 & 670
+def test_html_toc_with_custom_rendering(tmp_path):  # issue 240 & 670
     class PDF(FPDF):
         HTML2FPDF_CLASS = CustomHTML2FPDF
 
@@ -271,4 +271,4 @@ class PDF(FPDF):
         <h6>Level 6</h6>
         <p>paragraph</p>"""
         )
-    assert_pdf_equal(pdf, HERE / "custom_HTML2FPDF.pdf", tmp_path)
+    assert_pdf_equal(pdf, HERE / "html_toc_with_custom_rendering.pdf", tmp_path)