v0.6.0 BigUpdate

md2gost/office/uno_worker.py

@@ -11,7 +11,8 @@ import sys
 import time
 
 from uno_helpers import (
-    MAX_PASSES,
+    PASS_SAFETY_CAP,
+    continuation_pass_limit,
     CaptionInfo,
     as_unicode,
     caption_style_name,
@@ -22,6 +23,8 @@ from uno_helpers import (
     BREAK_PAGE_BEFORE,
     CONTROL_PARAGRAPH_BREAK,
     find_page_break_row,
+    locate_break_row,
+    resume_page_before_cut,
     is_caption_orphaned,
     is_orphan_header_break,
     parse_caption_text,
@@ -489,6 +492,20 @@ def _caption_for_split(doc, table):
     )
 
 
+def _count_split_targets(doc, tables, listings):
+    """Captioned tables and listings this pass is allowed to split."""
+    found = 0
+    for table in _body_tables(doc):
+        caption = _caption_before_table(doc, table)
+        if caption is None:
+            continue
+        if caption.kind == "table" and tables:
+            found += 1
+        elif caption.kind == "listing" and listings:
+            found += 1
+    return found
+
+
 def _caption_page_before_table(doc, table):
     try:
         anchor = _uno_call(table, "Anchor")
@@ -506,6 +523,26 @@ def _caption_page_before_table(doc, table):
         return None
 
 
+def _page_at_row(doc, table, row_1based):
+    """Page number of one row, or None. Rows are 1-based, UNO cells are 0-based."""
+    row = int(row_1based) - 1
+    if row < 0:
+        return None
+    try:
+        ncols = int(table.getColumns().getCount())
+    except Exception:
+        ncols = 1
+    for col in range(max(1, ncols)):
+        try:
+            cell = table.getCellByPosition(col, row)
+            page = _page_at_range(doc, _uno_call(cell, "Start"))
+            if page is not None:
+                return int(page)
+        except Exception:
+            continue
+    return None
+
+
 def _row_page_numbers(doc, table):
     pages = []
     try:
@@ -1049,10 +1086,16 @@ def _copy_header_to_continuation(first_table, cont_table):
         pass
 
 
-def _fix_pass(uno, ctx, doc, tables, listings, repeat_header, details):
+def _fix_pass(uno, ctx, doc, tables, listings, repeat_header, details, start_page=1):
+    """Split the first crossing table. Returns (made, cut_page)."""
     made = 0
+    cut_page = None
+    try:
+        start_page = int(start_page)
+    except (TypeError, ValueError):
+        start_page = 1
     body_tables = _body_tables(doc)
-    for table in reversed(body_tables):
+    for table in body_tables:
         caption = _caption_for_split(doc, table)
         if caption is None:
             continue
@@ -1061,9 +1104,18 @@ def _fix_pass(uno, ctx, doc, tables, listings, repeat_header, details):
         if caption.kind == "listing" and not listings:
             continue
 
-        pages = _row_page_numbers(doc, table)
-        if caption.kind == "table" and pages and not caption.is_continuation:
-            if is_caption_orphaned(_caption_page_before_table(doc, table), pages[0]):
+        try:
+            row_count = int(table.getRows().getCount())
+        except Exception:
+            continue
+        if start_page > 1 and row_count > 0:
+            last_page = _page_at_row(doc, table, row_count)
+            if last_page is not None and last_page < start_page:
+                continue
+
+        first_page = _page_at_row(doc, table, 1) if row_count else None
+        if caption.kind == "table" and first_page is not None and not caption.is_continuation:
+            if is_caption_orphaned(_caption_page_before_table(doc, table), first_page):
                 try:
                     if _page_break_before_table_caption(doc, table):
                         details.append(
@@ -1082,10 +1134,14 @@ def _fix_pass(uno, ctx, doc, tables, listings, repeat_header, details):
                     details.append("skip: orphan-caption page break: {0}".format(exc))
                 continue
 
-        break_at = find_page_break_row(pages)
+        break_at, cut_page = locate_break_row(
+            row_count,
+            lambda row, current=table: _page_at_row(doc, current, row),
+            start_page,
+        )
         if break_at is None:
             continue
-        if break_at > int(table.getRows().getCount()):
+        if break_at > row_count:
             continue
 
         had_header = _table_has_header_row(table)
@@ -1169,9 +1225,9 @@ def _fix_pass(uno, ctx, doc, tables, listings, repeat_header, details):
         details.append(
             "split: {0} {1} @row {2}".format(caption.kind, caption.number, break_at)
         )
-        return made
+        return made, cut_page
 
-    return made
+    return made, cut_page
 
 
 def run_worker(port, path, engine, tables, listings, repeat_header):
@@ -1187,14 +1243,30 @@ def run_worker(port, path, engine, tables, listings, repeat_header):
             doc, uno, ctx, tables, listings, details
         )
         _repair_split_table_borders(doc, tables, listings, details)
-        for _ in range(MAX_PASSES):
+        targets = _count_split_targets(doc, tables, listings)
+        passes_done = 0
+        start_page = 1
+        while passes_done < continuation_pass_limit(targets, splits):
+            if passes_done >= PASS_SAFETY_CAP:
+                details.append("skip: pass safety cap")
+                break
             _ensure_pagination(uno, ctx, doc)
-            made = _fix_pass(
-                uno, ctx, doc, tables, listings, repeat_header, details
+            made, cut_page = _fix_pass(
+                uno, ctx, doc, tables, listings, repeat_header, details, start_page
             )
+            passes_done += 1
             if made == 0:
+                if start_page > 1:
+                    start_page = 1
+                    details.append("recheck: from page 1")
+                    continue
                 break
             splits += made
+            if cut_page:
+                start_page = resume_page_before_cut(cut_page)
+                details.append(
+                    "resume: page {0} after cut at {1}".format(start_page, cut_page)
+                )
             doc = _reload_document(uno, ctx, doc, path, engine)
             _ensure_pagination(uno, ctx, doc)
         repaired += _repair_missing_continuations(