pyDiffTools 0.1.49__tar.gz → 0.1.50__tar.gz

{pydifftools-0.1.49/pyDiffTools.egg-info → pydifftools-0.1.50}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: pyDiffTools
-Version: 0.1.49
+Version: 0.1.50
 Summary: Diff tools
 Author: J M Franck
 License: Copyright (c) 2015, jmfranck
@@ -215,10 +215,21 @@ included are (listed in order of fun/utility):
 
     * this allows you to use an original file with good whitespace formatting as a "template" that you can match other (e.g. pandoc converted file) onto another
 
-    * changed prose lines exceeding 1.5 times the nearby reference width are
-      wrapped using the same sentence and punctuation rules as ``wr``.
-      Existing surrounding line breaks are retained; the fallback width is
-      80 columns when the reference provides too little prose.
+    * ``cpb`` and ``wmatch`` share a matcher that finds changed hunks, aligns
+      their words using Git's minimal diff, and restores reference whitespace.
+      It then considers extra breaks that preserve an unchanged reference line
+      without stranding a tiny edited fragment, and checks wrapping rules.
+
+    * ``--wrapnumber`` sets the maximum width (79 for ``cpb`` and ``wmatch``).
+      Wmatch also uses nearby reference widths for large edits, retaining
+      unchanged reference lines and protected Markdown blocks.
+
+    * ``--trailing-dependent-phrase`` defaults to 20: a clause boundary within
+      20 characters of the maximum width must end the line, even when its
+      trailing phrase would still fit. Commas, semicolons, colons, parentheses,
+      dashes, and inline-math boundaries are supported. Set it to 0 to disable
+      this rule. The option is shared by ``cpb``, ``wmatch``, ``wr`` and
+      ``wrchk``; ``--punctuation-slop`` remains an alias.
 
 - `pydifft wd` ("word diff"): generate "track changes" word files starting from pandoc markdown in a git history.  Assuming that you have copied diff-doc.js (copied + licensed from elsewhere) into your home directory, this will use pandoc to convert the markdown files to MS Word, then use the MS Word comparison tool to generate a document where all relevant changes are shown with "track changes."
 

{pydifftools-0.1.49 → pydifftools-0.1.50}/README.rst

@@ -160,10 +160,21 @@ included are (listed in order of fun/utility):
 
     * this allows you to use an original file with good whitespace formatting as a "template" that you can match other (e.g. pandoc converted file) onto another
 
-    * changed prose lines exceeding 1.5 times the nearby reference width are
-      wrapped using the same sentence and punctuation rules as ``wr``.
-      Existing surrounding line breaks are retained; the fallback width is
-      80 columns when the reference provides too little prose.
+    * ``cpb`` and ``wmatch`` share a matcher that finds changed hunks, aligns
+      their words using Git's minimal diff, and restores reference whitespace.
+      It then considers extra breaks that preserve an unchanged reference line
+      without stranding a tiny edited fragment, and checks wrapping rules.
+
+    * ``--wrapnumber`` sets the maximum width (79 for ``cpb`` and ``wmatch``).
+      Wmatch also uses nearby reference widths for large edits, retaining
+      unchanged reference lines and protected Markdown blocks.
+
+    * ``--trailing-dependent-phrase`` defaults to 20: a clause boundary within
+      20 characters of the maximum width must end the line, even when its
+      trailing phrase would still fit. Commas, semicolons, colons, parentheses,
+      dashes, and inline-math boundaries are supported. Set it to 0 to disable
+      this rule. The option is shared by ``cpb``, ``wmatch``, ``wr`` and
+      ``wrchk``; ``--punctuation-slop`` remains an alias.
 
 - `pydifft wd` ("word diff"): generate "track changes" word files starting from pandoc markdown in a git history.  Assuming that you have copied diff-doc.js (copied + licensed from elsewhere) into your home directory, this will use pandoc to convert the markdown files to MS Word, then use the MS Word comparison tool to generate a document where all relevant changes are shown with "track changes."
 

{pydifftools-0.1.49 → pydifftools-0.1.50/pyDiffTools.egg-info}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: pyDiffTools
-Version: 0.1.49
+Version: 0.1.50
 Summary: Diff tools
 Author: J M Franck
 License: Copyright (c) 2015, jmfranck
@@ -215,10 +215,21 @@ included are (listed in order of fun/utility):
 
     * this allows you to use an original file with good whitespace formatting as a "template" that you can match other (e.g. pandoc converted file) onto another
 
-    * changed prose lines exceeding 1.5 times the nearby reference width are
-      wrapped using the same sentence and punctuation rules as ``wr``.
-      Existing surrounding line breaks are retained; the fallback width is
-      80 columns when the reference provides too little prose.
+    * ``cpb`` and ``wmatch`` share a matcher that finds changed hunks, aligns
+      their words using Git's minimal diff, and restores reference whitespace.
+      It then considers extra breaks that preserve an unchanged reference line
+      without stranding a tiny edited fragment, and checks wrapping rules.
+
+    * ``--wrapnumber`` sets the maximum width (79 for ``cpb`` and ``wmatch``).
+      Wmatch also uses nearby reference widths for large edits, retaining
+      unchanged reference lines and protected Markdown blocks.
+
+    * ``--trailing-dependent-phrase`` defaults to 20: a clause boundary within
+      20 characters of the maximum width must end the line, even when its
+      trailing phrase would still fit. Commas, semicolons, colons, parentheses,
+      dashes, and inline-math boundaries are supported. Set it to 0 to disable
+      this rule. The option is shared by ``cpb``, ``wmatch``, ``wr`` and
+      ``wrchk``; ``--punctuation-slop`` remains an alias.
 
 - `pydifft wd` ("word diff"): generate "track changes" word files starting from pandoc markdown in a git history.  Assuming that you have copied diff-doc.js (copied + licensed from elsewhere) into your home directory, this will use pandoc to convert the markdown files to MS Word, then use the MS Word comparison tool to generate a document where all relevant changes are shown with "track changes."
 

{pydifftools-0.1.49 → pydifftools-0.1.50}/pyDiffTools.egg-info/SOURCES.txt

@@ -70,6 +70,7 @@ pydifftools/git_gd_image.py
 pydifftools/git_gd_qt.py
 pydifftools/html_comments.py
 pydifftools/html_uncomments.py
+pydifftools/line_alignment.py
 pydifftools/log_example.py
 pydifftools/match_spaces.py
 pydifftools/onewordify.py
@@ -83,6 +84,7 @@ pydifftools/split_conflict.py
 pydifftools/unseparate_comments.py
 pydifftools/update_check.py
 pydifftools/wrap_sentences.py
+pydifftools/wrapping_options.py
 pydifftools/xml2xlsx.vbs
 pydifftools/zotero.py
 pydifftools/flowchart/__init__.py
@@ -107,6 +109,7 @@ tests/test_forward_search.py
 tests/test_git_gd_history.py
 tests/test_git_gd_image.py
 tests/test_git_gd_qt.py
+tests/test_line_alignment.py
 tests/test_rrng.py
 tests/test_source_jump.py
 tests/test_tex_to_qmd.py

{pydifftools-0.1.49 → pydifftools-0.1.50}/pydifftools/browser_lifecycle.py

@@ -33,11 +33,40 @@ def launch_chrome(webdriver, **kwargs):
             "do not match; retrying with the driver version check disabled.",
             file=sys.stderr,
         )
-        kwargs["service"] = Service(service_args=["--disable-build-check"])
+        service = kwargs.get("service")
+        kwargs["service"] = Service(
+            executable_path=service.path if service else None,
+            service_args=[
+                *(service.service_args if service else []),
+                "--disable-build-check",
+            ],
+        )
         return webdriver.Chrome(**kwargs)
         # }}}
 
 
+def start_chrome():
+    """Use an installed driver without an online Selenium Manager lookup."""
+    # Keep Selenium imports lazy, as in the watch commands.
+    from selenium import webdriver
+    from selenium.webdriver.chrome.service import Service
+
+    driver_path = shutil.which("chromedriver")
+    if driver_path:
+        options = webdriver.ChromeOptions()
+        # Debian's driver defaults to Chromium even if Chrome is installed;
+        # retain Selenium Manager's preference for Google Chrome.
+        browser_path = shutil.which("google-chrome") or shutil.which(
+            "google-chrome-stable"
+        )
+        if browser_path:
+            options.binary_location = browser_path
+        return launch_chrome(
+            webdriver, service=Service(driver_path), options=options
+        )
+    return launch_chrome(webdriver)
+
+
 def prepare_for_dialog():
     """Let watch commands close their preview before a dialog appears."""
     for callback in list(dialog_callbacks):

{pydifftools-0.1.49 → pydifftools-0.1.50}/pydifftools/command_line.py

@@ -11,6 +11,7 @@ import nbformat
 import difflib
 import shutil
 import importlib.util
+import importlib.metadata
 from pathlib import Path
 from . import (
     match_spaces,
@@ -46,6 +47,9 @@ from .notebook.fast_build import (
 )
 
 from .command_registry import _COMMAND_SPECS, register_command
+from .wrapping_options import (
+    DEFAULT_WIDTH, DEFAULT_TRAILING_DEPENDENT_PHRASE, WRAPPING_ARGUMENTS,
+)
 
 _ARGCOMPLETE_SPEC = importlib.util.find_spec("argcomplete")
 if (
@@ -366,9 +370,19 @@ def gvr(arguments):
 @register_command(
     "match whitespace",
     filename_extensions={"arguments": [".md", ".tex"]},
+    argument_options={
+        **WRAPPING_ARGUMENTS,
+        "arguments": {
+            "nargs": 2, "metavar": ("REFERENCE", "FILE"),
+            "help": "Reference file and current file to reflow.",
+        },
+    },
 )
-def wmatch(arguments):
-    match_spaces.run(arguments)
+def wmatch(
+    arguments, wrapnumber=DEFAULT_WIDTH,
+    trailing_dependent_phrase=DEFAULT_TRAILING_DEPENDENT_PHRASE,
+):
+    match_spaces.run(arguments, wrapnumber, trailing_dependent_phrase)
 
 
 @register_command("split conflict")
@@ -879,6 +893,12 @@ def build_parser():
         formatter_class=argparse.RawDescriptionHelpFormatter,
         is_root_parser=True,
     )
+    parser.add_argument(
+        "--ver",
+        "--version",
+        action="version",
+        version=importlib.metadata.version("pyDiffTools"),
+    )
     parser.epilog = _subcommand_help_hint(parser.prog)
     parser._pydifft_subparsers = {}
     subparsers = parser.add_subparsers(dest="command")
@@ -913,6 +933,9 @@ def build_parser():
 def main(argv=None):
     if argv is None:
         argv = sys.argv[1:]
+    if argv and argv[0] in ("--ver", "--version"):
+        build_parser().parse_args(argv)
+        return
     # Run the PyPI update check once per UTC day so users see a notice but
     # startup stays fast when offline. The date is stored in the
     # PYDIFFTOOLS_UPDATE_CHECK_LAST_RAN_UTC_DATE environment variable.
@@ -959,6 +982,16 @@ def main(argv=None):
             parser._pydifft_subparsers[subcommand].print_help()
             return
     namespace = parser.parse_args(argv)
+    # Shared wrapping arguments are validated identically for each command.
+    if hasattr(namespace, "wrapnumber") and namespace.wrapnumber < 1:
+        parser._pydifft_subparsers[namespace.command].error(
+            "--wrapnumber must be at least 1"
+        )
+    if getattr(namespace, "trailing_dependent_phrase",
+               getattr(namespace, "punctuation_slop", 0)) < 0:
+        parser._pydifft_subparsers[namespace.command].error(
+            "--trailing-dependent-phrase must be nonnegative"
+        )
     if (
         namespace.command == "cpb"
         and namespace.comments_to_margin

{pydifftools-0.1.49 → pydifftools-0.1.50}/pydifftools/command_registry.py

@@ -15,6 +15,7 @@ def register_command(
     description=None,
     help=None,
     filename_extensions=None,
+    argument_options=None,
 ):
     """Register a command handler for the CLI dispatcher."""
 
@@ -108,6 +109,12 @@ def register_command(
                     kwargs["type"] = type(parameter.default)
             if parameter.name in argument_help:
                 kwargs["help"] = argument_help[parameter.name].strip()
+            if argument_options is not None:
+                kwargs.update(argument_options.get(parameter.name, {}))
+                override_flags = kwargs.pop("flags", None)
+                if override_flags is not None:
+                    flags = override_flags
+                    kwargs["dest"] = parameter.name
             argument_spec = {"flags": flags, "kwargs": kwargs}
             if parameter.name in completion_allowednames:
                 argument_spec["completion_allowednames"] = (

{pydifftools-0.1.49 → pydifftools-0.1.50}/pydifftools/comment_tags.lua

@@ -237,9 +237,13 @@ local function detect_block_opener(block)
   end
 
   if para_like_t(block) then
-    for k, opener_spec in pairs(OPENERS) do
-      local found, before, after = split_inline_block_at_tag(block, k)
-      if found then
+    -- Scan the paragraph once, regardless of the number of authors.
+    local content = block.content
+    for index = 1, #content do
+      local tag = raw_inline_html(content[index])
+      local opener_spec = tag and OPENERS[tag] or nil
+      if opener_spec then
+        local _, before, after = split_inline_block_at_tag(block, tag)
         return opener_spec, before, after
       end
     end
@@ -452,3 +456,6 @@ function Pandoc(doc)
   end
   return doc:walk({ Inlines = Inlines, Blocks = Blocks, Div = Div })
 end
+
+-- Initialize authors before walking comments, without an implicit first walk.
+return {{ Pandoc = Pandoc }}

{pydifftools-0.1.49 → pydifftools-0.1.50}/pydifftools/comment_tags_margin.lua

@@ -273,9 +273,13 @@ local function detect_block_opener(block)
   end
 
   if para_like_t(block) then
-    for k, opener_spec in pairs(OPENERS) do
-      local found, before, after = split_inline_block_at_tag(block, k)
-      if found then
+    -- Scan the paragraph once, regardless of the number of authors.
+    local content = block.content
+    for index = 1, #content do
+      local tag = raw_inline_html(content[index])
+      local opener_spec = tag and OPENERS[tag] or nil
+      if opener_spec then
+        local _, before, after = split_inline_block_at_tag(block, tag)
         return opener_spec, before, after
       end
     end
@@ -487,3 +491,6 @@ function Pandoc(doc)
   end
   return doc:walk({ Inlines = Inlines, Blocks = Blocks, Div = Div })
 end
+
+-- Initialize authors before walking comments, without an implicit first walk.
+return {{ Pandoc = Pandoc }}

{pydifftools-0.1.49 → pydifftools-0.1.50}/pydifftools/comment_tags_no_comments.lua

@@ -166,9 +166,13 @@ local function detect_block_opener(block)
   end
 
   if para_like_t(block) then
-    for k, opener_spec in pairs(OPENERS) do
-      local found, before, after = split_inline_block_at_tag(block, k)
-      if found then
+    -- Scan the paragraph once, regardless of the number of authors.
+    local content = block.content
+    for index = 1, #content do
+      local tag = raw_inline_html(content[index])
+      local opener_spec = tag and OPENERS[tag] or nil
+      if opener_spec then
+        local _, before, after = split_inline_block_at_tag(block, tag)
         return opener_spec, before, after
       end
     end
@@ -348,3 +352,6 @@ function Pandoc(doc)
   end
   return doc:walk({ Inlines = Inlines, Blocks = Blocks, Div = Div })
 end
+
+-- Initialize authors before walking comments, without an implicit first walk.
+return {{ Pandoc = Pandoc }}

{pydifftools-0.1.49 → pydifftools-0.1.50}/pydifftools/continuous.py

@@ -18,6 +18,9 @@ from watchdog.events import FileSystemEventHandler
 from watchdog.observers import Observer
 from watchdog.observers.polling import PollingObserver
 from .command_registry import register_command
+from .wrapping_options import (
+    DEFAULT_WIDTH, DEFAULT_TRAILING_DEPENDENT_PHRASE, WRAPPING_ARGUMENTS,
+)
 from .zotero import recover_bibliography, zotero_notice
 from .comment_migration import prepare_comment_source
 from .browser_lifecycle import (
@@ -25,8 +28,8 @@ from .browser_lifecycle import (
     close_browser_window,
     dialog_callbacks,
     forward_search_in_browser,
-    launch_chrome,
     prepare_for_dialog,
+    start_chrome,
 )
 from .forward_search import (
     CPB_FORWARD_SEARCH_PORT,
@@ -170,6 +173,54 @@ def show_markdown_fix_dialog(report):
     dialog._pydifftools_application = application
     dialog.setWindowTitle("Markdown source fixes")
     fixed_font = QFontDatabase.systemFont(QFontDatabase.SystemFont.FixedFont)
+
+    def annotated_source(before, after, line):
+        """Mark changed breaks on their positions in the original source."""
+        original_words = list(re.finditer(r"\S+", before))
+        updated_words = list(re.finditer(r"\S+", after))
+        if [word[0] for word in original_words] != [
+            word[0] for word in updated_words
+        ]:
+            return None
+        marks = {}
+        for number in range(len(original_words) + 1):
+            start = original_words[number - 1].end() if number else 0
+            stop = (
+                original_words[number].start()
+                if number < len(original_words)
+                else len(before)
+            )
+            updated_start = updated_words[number - 1].end() if number else 0
+            updated_stop = (
+                updated_words[number].start()
+                if number < len(updated_words)
+                else len(after)
+            )
+            old_breaks = [
+                start + match.start()
+                for match in re.finditer("\n", before[start:stop])
+            ]
+            new_count = after[updated_start:updated_stop].count("\n")
+            if new_count > len(old_breaks):
+                marks[start] = (
+                    "<span style='color:#188038; font-weight:bold'>↳</span>"
+                    * (new_count - len(old_breaks))
+                )
+            for position in old_breaks[new_count:]:
+                marks[position] = (
+                    "<span style='color:#c62828; font-weight:bold'>↳×</span>"
+                )
+        if not marks:
+            return None
+        marked = "".join(
+            marks.get(position, "") + escape(char)
+            for position, char in enumerate(before)
+        ) + marks.get(len(before), "")
+        return "<br>".join(
+            f"<span style='color:#666'>{line + offset:>6}</span> " + value
+            for offset, value in enumerate(marked.split("\n"))
+        )
+
     # {{{ group fixes of one kind onto pages that fill the screen
     screen_height = dialog.screen().availableGeometry().height()
     dialog.resize(900, int(screen_height * 0.85))
@@ -189,13 +240,15 @@ def show_markdown_fix_dialog(report):
         page = []
         rows = 0
         for fix in sorted(group, key=lambda fix: fix["line"]):
-            # a blank line separates each fix from the one before it, and a
-            # fix that only drops trailing spaces shows just its old line
+            # A break annotation needs only the original source lines.
             fix_rows = (
                 len(fix["before"].splitlines() or [""])
                 + (
                     0
                     if fix["after"] == fix["before"].rstrip()
+                    or annotated_source(
+                        fix["before"], fix["after"], fix["line"]
+                    ) is not None
                     else len(fix["after"].splitlines() or [""])
                 )
                 + (1 if page else 0)
@@ -235,6 +288,7 @@ def show_markdown_fix_dialog(report):
     preview = QTextEdit()
     preview.setReadOnly(True)
     preview.setFont(fixed_font)
+    preview.setLineWrapMode(QTextEdit.LineWrapMode.NoWrap)
     preview_row.addWidget(line_numbers)
     preview_row.addWidget(preview, 1)
     layout.addLayout(preview_row, 1)
@@ -261,9 +315,17 @@ def show_markdown_fix_dialog(report):
             )
             details.setText(record["reason"])
             line_numbers.setVisible(False)
-            # {{{ show each fix as numbered before (−) and after (+) lines
+            # {{{ annotate changed breaks on the original source
             entries = []
+            has_break_annotations = False
             for fix in record["fixes"]:
+                annotated = annotated_source(
+                    fix["before"], fix["after"], fix["line"]
+                )
+                if annotated is not None:
+                    entries.append(annotated)
+                    has_break_annotations = True
+                    continue
                 rows = []
                 for offset, value in enumerate(
                     fix["before"].splitlines() or [""]
@@ -289,6 +351,13 @@ def show_markdown_fix_dialog(report):
                         + escape(value)
                     )
                 entries.append("<br>".join(rows))
+            if has_break_annotations:
+                details.setText(
+                    record["reason"]
+                    + "<br>Original source: "
+                    "<span style='color:#188038'>↳</span> inserted newline; "
+                    "<span style='color:#c62828'>↳×</span> removed newline."
+                )
             preview.setHtml(
                 "<pre style='font-family:monospace; margin:0'>"
                 + "<br><br>".join(entries)
@@ -308,21 +377,33 @@ def show_markdown_fix_dialog(report):
                 "lay them out as suggested."
             )
             line_numbers.setVisible(False)
-            # {{{ show HEAD, the current source and the suggested layout
+            # {{{ show suggested breaks on the original source
             sections = []
-            for title, text, numbered in (
-                ("git HEAD:", record["head"], False),
-                ("Now:", record["current"], True),
-                ("Suggested:", record["suggested"], True),
-            ):
-                rows = [escape(title)]
-                for offset, value in enumerate(text.split("\n")):
-                    number = record["line"] + offset if numbered else ""
-                    rows.append(
-                        f"<span style='color:#666'>{number:>6}</span> "
-                        + escape(value)
-                    )
-                sections.append("<br>".join(rows))
+            annotated = annotated_source(
+                record["current"], record["suggested"], record["line"]
+            )
+            if annotated is not None:
+                sections.append(annotated)
+                details.setText(
+                    details.text()
+                    + "<br>Original source, with suggested changes: "
+                    "<span style='color:#188038'>↳</span> insert newline; "
+                    "<span style='color:#c62828'>↳×</span> remove newline."
+                )
+            else:
+                for title, text, numbered in (
+                    ("git HEAD:", record["head"], False),
+                    ("Now:", record["current"], True),
+                    ("Suggested:", record["suggested"], True),
+                ):
+                    rows = [escape(title)]
+                    for offset, value in enumerate(text.split("\n")):
+                        number = record["line"] + offset if numbered else ""
+                        rows.append(
+                            f"<span style='color:#666'>{number:>6}</span> "
+                            + escape(value)
+                        )
+                    sections.append("<br>".join(rows))
             preview.setHtml(
                 "<pre style='font-family:monospace; margin:0'>"
                 + "<br><br>".join(sections)
@@ -415,7 +496,8 @@ def run_pandoc(
     comments_to_margin=False,
     no_comments=False,
     comment_filter_session=None,
-    wrapnumber=79,
+    wrapnumber=DEFAULT_WIDTH,
+    trailing_dependent_phrase=DEFAULT_TRAILING_DEPENDENT_PHRASE,
 ):
     # {{{ automatically fix Markdown source and request edits for
     # unfinished spans
@@ -424,7 +506,8 @@ def run_pandoc(
     automatic_fixes_were_made = False
     while True:
         report = autofix_markdown_file(
-            filename, wrapnumber=wrapnumber, git_head=True
+            filename, wrapnumber=wrapnumber, git_head=True,
+            punctuation_slop=trailing_dependent_phrase,
         )
         automatic_fixes_were_made |= bool(report["fixes"])
         if report["fixes"] or report["warnings"] or report["layout"]:
@@ -927,6 +1010,8 @@ class Handler(FileSystemEventHandler):
             self._queue_if_source_changed(event.src_path, event.dest_path)
 
 
+# also used by: command_line.main through the command registry, and the
+# CPB runtime tests in tests/test_continuous_shutdown.py and test_zotero.py.
 @register_command(
     "continuous pandoc build.  Like latexmk, but for markdown!",
     help={
@@ -938,14 +1023,15 @@ class Handler(FileSystemEventHandler):
         "no_comments": (
             "Render the HTML without comment tags or comment div blocks."
         ),
-        "wrapnumber": (
-            "Maximum source line width; the automatic Markdown fixes break "
-            "only lines longer than this."
-        ),
     },
     filename_extensions={"filename": ".md"},
+    argument_options=WRAPPING_ARGUMENTS,
 )
-def cpb(filename, comments_to_margin=False, no_comments=False, wrapnumber=79):
+def cpb(
+    filename, comments_to_margin=False, no_comments=False,
+    wrapnumber=DEFAULT_WIDTH,
+    trailing_dependent_phrase=DEFAULT_TRAILING_DEPENDENT_PHRASE,
+):
     source_path = os.path.normpath(os.path.abspath(filename))
     source_dir = os.path.dirname(source_path)
     html_file = filename.rsplit(".", 1)[0] + ".html"
@@ -961,6 +1047,7 @@ def cpb(filename, comments_to_margin=False, no_comments=False, wrapnumber=79):
         daemon=True,
     )
     chrome = None
+    scroll_position = 0
     observer = None
     observer_started = False
     socket_thread_started = False
@@ -993,6 +1080,7 @@ def cpb(filename, comments_to_margin=False, no_comments=False, wrapnumber=79):
             no_comments=no_comments,
             comment_filter_session=comment_filter_session,
             wrapnumber=wrapnumber,
+            trailing_dependent_phrase=trailing_dependent_phrase,
         )
         source_jump_server = SourceJumpServer(source_path)
         source_jump_server.start()
@@ -1000,10 +1088,9 @@ def cpb(filename, comments_to_margin=False, no_comments=False, wrapnumber=79):
 
         # Selenium is deliberately initialized once, after the first
         # successful build. Nothing in recovery constructs a browser.
-        from selenium import webdriver
         from selenium.common.exceptions import WebDriverException
 
-        chrome = launch_chrome(webdriver)
+        chrome = start_chrome()
         observer = Observer()
         change_queue = queue.Queue()
         event_handler = Handler(filename, change_queue)
@@ -1052,8 +1139,12 @@ def cpb(filename, comments_to_margin=False, no_comments=False, wrapnumber=79):
         def close_for_dialog():
             # a rebuild needs the user, so close the preview while the
             # issues are dealt with; it reopens once the rebuild is done
-            nonlocal chrome
+            nonlocal chrome, scroll_position
             if chrome is not None:
+                # sessionStorage does not survive quitting Chrome.
+                scroll_position = chrome.execute_script(
+                    "return window.scrollY;"
+                )
                 close_browser_window(chrome)
                 chrome = None
 
@@ -1140,6 +1231,9 @@ def cpb(filename, comments_to_margin=False, no_comments=False, wrapnumber=79):
                                 no_comments=no_comments,
                                 comment_filter_session=comment_filter_session,
                                 wrapnumber=wrapnumber,
+                                trailing_dependent_phrase=(
+                                    trailing_dependent_phrase
+                                ),
                             )
                             append_autorefresh(
                                 html_file, source_jump_server.url
@@ -1185,10 +1279,14 @@ def cpb(filename, comments_to_margin=False, no_comments=False, wrapnumber=79):
                                     break
                         if chrome is None:
                             # {{{ reopen the preview a dialog closed
-                            chrome = launch_chrome(webdriver)
+                            chrome = start_chrome()
                             chrome.get(
                                 "file://" + os.path.abspath(html_file)
                             )
+                            chrome.execute_script(
+                                "window.scrollTo(0, arguments[0]);",
+                                scroll_position,
+                            )
                             # }}}
             # }}}
             time.sleep(MAIN_LOOP_INTERVAL_SECONDS)