Clear the last mypy errors and store the resolvable DOI

Closes #18.

extract_cv.py, the one finding that was a real shape problem rather than
a stub gap: preprocess_content returned a bare string when
extract_footnotes was false and a (footnote, content) pair when it was
true. There is exactly one caller in the repo and it always passes true,
so the string branch was dead code that its own docstring described as
backwards compatibility. Dropped the flag and the function now always
returns the pair. Losing the else branch also loses its blfootnote
removal, which is equivalent: extract_footnote strips the footnote it
extracts, and when there is no footnote there is nothing to strip.

extract_people.py gets one attr_str helper rather than a cast per call
site. BeautifulSoup types attribute access as str, AttributeValueList or
None because a multi-valued attribute like class comes back as a list;
src and href never do, so the helper joins a list and passes a string
through. It applies at six sites, not the three mypy first reported --
fixing those unmasked two more that had been hidden behind them.

onboard_member.py: the RGBA corner check now tests isinstance before
indexing, which behaves as before for a non-tuple since the enclosing
except already caught that; and the re.subn lambda became a named
function so mypy can pin the AnyStr overload.

The publications DOI goes back to 10.31234/osf.io/mhxtd_v1, which
resolves. PsyArXiv has not minted the unversioned form, so the shorter
one was a dead identifier. The column is still never rendered.

Verified by rebuilding: JRM_CV.html is byte-identical after the parser
change and both section-note footnotes still render, which is what
actually proves it, and all four content pages are unchanged too.
extract_people.py has no tests, so its extraction was exercised directly
against the live page: 181 rows over 7 sections with every url and image
field a string. mypy and ruff are both clean and the suite is 348 passed,
12 skipped.

Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01QQUWxAFYESZCZe3DLyy1eB

Author: jeremymanning

README.md

@@ -456,16 +456,22 @@ The `extract_cv.py` script provides a custom LaTeX-to-HTML converter that handle
 ### Key Functions in extract_cv.py
 
 | Function | Purpose |
-|----------|---------|
+|-|-|
 | `extract_document_body()` | Extract content between `\begin{document}` and `\end{document}` |
 | `balanced_braces_extract()` | Parse nested LaTeX braces correctly |
 | `convert_latex_formatting()` | Convert LaTeX commands to HTML |
 | `parse_etaremune()` | Parse reverse-numbered publication lists |
 | `extract_header_info()` | Extract name and contact information |
 | `extract_sections()` | Split document into sections/subsections |
+| `extract_footnote()` | Pull `\blfootnote{}` out of a section, returning `(footnote_html, remaining_content)` |
+| `preprocess_content()` | Strip comments, the footnote, `\vspace`, and the "Last updated" line before parsing; returns `(footnote, cleaned_content)` |
 | `render_section_content()` | Convert section content based on type |
 | `generate_html()` | Assemble complete HTML document |
 
+> `preprocess_content()` is the single entry point for footnote handling — it always
+> returns the `(footnote, content)` pair, and `extract_footnote()` strips the
+> `\blfootnote` it extracts, so a footnote never survives into the rendered body.
+
 ### CV Stylesheet (cv.css)
 
 The stylesheet provides:

data/publications.xlsx

@@ -8,7 +8,7 @@
 
 import re
 from pathlib import Path
-from typing import Dict, List
+from typing import Any, Dict, List, Optional, Tuple
 from dataclasses import dataclass, field
 
 
@@ -280,7 +280,7 @@ def parse_labeled_lists(content: str) -> List[tuple]:
 
 def extract_header_info(body: str) -> Dict[str, str]:
     """Extract header information (name, title, contact)."""
-    info = {}
+    info: Dict[str, Any] = {}
 
     # Find header section (before first section)
     header_match = re.search(r'^(.+?)\\section\*', body, re.DOTALL)
@@ -428,7 +428,7 @@ def render_labeled_lists(labeled_lists: List[tuple], use_two_column: bool = Fals
     return html
 
 
-def extract_footnote(content: str) -> tuple:
+def extract_footnote(content: str) -> Tuple[Optional[str], str]:
     """Extract blfootnote content from text. Returns (footnote_text, cleaned_content)."""
     pattern = r'\\blfootnote\{'
     match = re.search(pattern, content)
@@ -446,22 +446,17 @@ def extract_footnote(content: str) -> tuple:
     return None, content
 
 
-def preprocess_content(content: str, extract_footnotes: bool = False) -> tuple:
+def preprocess_content(content: str) -> Tuple[Optional[str], str]:
     """Preprocess content to handle problematic LaTeX commands before parsing.
 
-    If extract_footnotes is True, returns (footnote, cleaned_content).
-    Otherwise returns just cleaned_content for backwards compatibility.
+    Returns (footnote, cleaned_content). extract_footnote strips the
+    \\blfootnote it extracts, so the footnote never survives into the body.
     """
     # Remove comments first
     content = re.sub(r'^%.*$', '', content, flags=re.MULTILINE)
     content = re.sub(r'(?<!\\)%.*$', '', content, flags=re.MULTILINE)
 
-    footnote = None
-    if extract_footnotes:
-        footnote, content = extract_footnote(content)
-    else:
-        # Remove blfootnote (with nested braces)
-        content = remove_command_with_braces(content, 'blfootnote')
+    footnote, content = extract_footnote(content)
 
     # Remove vspace
     content = remove_command_with_braces(content, 'vspace')
@@ -470,16 +465,14 @@ def preprocess_content(content: str, extract_footnotes: bool = False) -> tuple:
     content = re.sub(r'\{\\scriptsize\s*Last updated:.*?\\today\s*\}', '', content, flags=re.DOTALL)
     content = re.sub(r'Last updated:\s*$', '', content, flags=re.MULTILINE)
 
-    if extract_footnotes:
-        return footnote, content
-    return content
+    return footnote, content
 
 
 def render_section_content(content: str, section_title: str) -> str:
     """Render section content to HTML based on section type."""
 
     # Extract footnote first (for display as note under section header)
-    footnote, content = preprocess_content(content, extract_footnotes=True)
+    footnote, content = preprocess_content(content)
 
     # Build HTML with optional footnote note
     html_prefix = ''

scripts/extract_cv.py

@@ -33,7 +33,7 @@ def extract_people(html_path: Path) -> dict:
         role = name_parts[1].strip() if len(name_parts) > 1 else ''
 
         data['director'] = [{
-            'image': img.get('src', '').replace('images/people/', '') if img else '',
+            'image': attr_str(img, 'src').replace('images/people/', '') if img else '',
             'name': name,
             'role': role,
             'bio': ps[0].get_text(strip=True) if ps else '',
@@ -95,7 +95,7 @@ def extract_people(html_path: Path) -> dict:
             if link and p.get_text(strip=True):
                 collab = {
                     'name': link.get_text(strip=True),
-                    'url': link.get('href', ''),
+                    'url': attr_str(link, 'href'),
                     'description': p.get_text(strip=True)
                 }
                 collaborators.append(collab)
@@ -120,7 +120,7 @@ def extract_person_card(card) -> dict:
     if h3:
         link = h3.find('a')
         if link:
-            name_url = link.get('href', '')
+            name_url = attr_str(link, 'href')
 
         text = h3.get_text()
         if '|' in text:
@@ -131,7 +131,7 @@ def extract_person_card(card) -> dict:
             name = text.strip()
 
     return {
-        'image': img.get('src', '').replace('images/people/', '') if img else '',
+        'image': attr_str(img, 'src').replace('images/people/', '') if img else '',
         'name': name,
         'name_url': name_url,
         'role': role,
@@ -204,7 +204,7 @@ def extract_alumni_list(elem) -> list:
         # Get name - either from link or from text before parenthesis
         if name_link:
             alum['name'] = name_link.get_text(strip=True)
-            alum['name_url'] = name_link.get('href', '')
+            alum['name_url'] = attr_str(name_link, 'href')
         else:
             # No name link, name is text before (
             if paren_pos > 0:
@@ -214,7 +214,7 @@ def extract_alumni_list(elem) -> list:
 
         # Get position link URL if available
         if position_link:
-            alum['current_position_url'] = position_link.get('href', '')
+            alum['current_position_url'] = attr_str(position_link, 'href')
 
         # Parse years and current position from parenthesis
         paren_match = re.search(r'\(([^)]+)\)', full_text)
@@ -268,6 +268,20 @@ def extract_alumni_simple_list(elem) -> list:
     return alumni
 
 
+def attr_str(tag: Any, name: str) -> str:
+    """Read a single-valued tag attribute as a string.
+
+    BeautifulSoup types attribute access as str | AttributeValueList | None,
+    because a multi-valued attribute such as class comes back as a list. The
+    attributes read here (src, href) are always single-valued, so join a list
+    rather than letting one leak into a field declared as text.
+    """
+    value = tag.get(name, '')
+    if isinstance(value, list):
+        return ' '.join(value)
+    return value or ''
+
+
 def get_inner_html(element) -> str:
     """Get the inner HTML of an element as a string."""
     if not element:

scripts/extract_people.py

@@ -640,8 +640,10 @@ def photo_already_processed(photo_base: str, project_root: Path) -> bool:
             img.getpixel((0, h - 1)),
             img.getpixel((w - 1, h - 1)),
         ]
-        # All corners should be fully transparent (alpha == 0)
-        if not all(c[3] == 0 for c in corners):
+        # All corners should be fully transparent (alpha == 0). getpixel is
+        # typed as float | tuple | None for single-band images; mode is
+        # RGBA here, so each corner is a 4-tuple.
+        if not all(isinstance(c, tuple) and c[3] == 0 for c in corners):
             return False
 
         return True
@@ -1040,9 +1042,11 @@ def close_cv_entry(cv_path: Path, name: str, end_year: str) -> bool:
     # count=1: close the first open entry only. Closing every one at a stroke
     # would rewrite an unrelated second open entry for the same name, which is
     # never what a single role change means.
+    def close_range(m: "re.Match[str]") -> str:
+        return f"{m.group(1)}{end_year})"
+
     new_content, count = re.subn(
-        pattern, lambda m: f"{m.group(1)}{end_year})", content, count=1,
-        flags=re.IGNORECASE
+        pattern, close_range, content, count=1, flags=re.IGNORECASE
     )
     if count == 0:
         return False