literate-programming

name

literate-programming

description

CRITICAL: ALWAYS activate this skill BEFORE making ANY changes to .nw files. Use proactively when: (1) creating, editing, reviewing, or improving any .nw file, (2) planning to add/modify functionality in files with .nw extension, (3) user asks about literate quality, (4) user mentions noweb, literate programming, tangling, or weaving, (5) working in directories containing .nw files, (6) creating new modules/files that will be .nw format. Trigger phrases: 'create module', 'add feature', 'update', 'modify', 'fix' + any .nw file. Never edit .nw files directly without first activating this skill to ensure literate programming principles are applied. (project, gitignored)

Literate Programming Skill

CRITICAL: This skill MUST be activated BEFORE making any changes to .nw files!

You are an expert in literate programming using the noweb system.

Reference Files

This skill includes detailed references in references/:

File	Content	Search patterns
noweb-commands.md	Tangling, weaving, flags, troubleshooting	notangle, noweave, -R, -L
testing-patterns.md	Test organization, placement, dependency testing	test functions, pytest, after implementation
overview-patterns.md	Whole-picture overviews, roadmap prose, structural diagrams	overview, diagram, tikz, roadmap
git-workflow.md	Version control, .gitignore, pre-commit	git, commit, generated files
multi-directory-projects.md	Large project organization, makefiles	src/, doc/, tests/, MODULES
project-initialization.md	New project setup, templates, checklist	new project, initialize, pyproject.toml
preamble.tex	Standard LaTeX preamble for documentation	\usepackage, memoir
build-pythontex-biber.md	Build gotchas: minted v3, the latexmkrc/PythonTeX mylatex fudge (and that latexmkrc is tangled from tex.mk.nw → gitignore it), PythonTeX dirs writing outside the repo, biber-vs-bibtex .bbl routing under -use-make, regenerating didactic.sty from the submodule	latexmkrc, pythontex, bibtexu, didactic.sty, -use-make

When to Use This Skill

Correct Workflow

1. User asks to modify a .nw file
2. YOU ACTIVATE THIS SKILL IMMEDIATELY
3. You plan the changes with literate programming principles
4. You make the changes following the principles
5. You regenerate code with make/notangle

Anti-pattern (NEVER do this)

1. User asks to modify a .nw file
2. You directly edit the .nw file ← WRONG
3. Later review finds literate quality problems
4. You have to redo everything

Remember

• .nw files are NOT regular source code files
• They combine documentation and code for human readers
• Literate quality is AS IMPORTANT as code correctness
• Bad literate quality = failed task, even if code works

Planning Changes

When making changes to a .nw file:

1. Read the existing file to understand structure and narrative
2. Plan with literate programming in mind:
   • What is the "why" behind this change?
   • Why does this approach work, not just why was it chosen?
   • How does this fit into the existing narrative?
   • Does the document need an early whole-picture overview before details?
   • Which overview, intro, roadmap figure, or relevant README.md summarizes this area, and does this change make it stale?
   • What new chunks are needed? What are their meaningful names?
   • Where in the pedagogical order should this be explained?
3. Design documentation BEFORE writing code:
   • Write prose explaining the problem and solution
   • Use subsections to structure complex explanations
   • For substantial .nw files, give maintainers the whole before the parts. If relationships are hard to hold in prose, add a structural figure. See references/overview-patterns.md.
4. Decompose code into well-named chunks:
   • Each chunk = one coherent concept
   • Names describe purpose, not syntax (like pseudocode)
5. Write the code chunks
6. Regenerate and test

Key principle: If you find yourself writing code comments to explain logic, that explanation belongs in the documentation chunks instead.

Reviewing Literate Programs

When reviewing, evaluate:

1. Narrative flow: Coherent story? Pedagogical order?
2. Variation theory: Contrasts used? "Whole, parts, whole" structure? Maintainer-facing .nw files should orient the reader to the system, then drill into parts, then reconnect changes to the whole.
3. Chunk quality: Meaningful names? Focused on single concepts? Bucket chunks named with nouns (<<option completions>>), not verb phrases (<<wire option completions>>)?
4. Explanation quality: Explains "why" not just "what"? The explanation should also say why the chosen approach works. Red flags: prose that begins "We [verb] the [noun]" matching a function name; prose that describes parameter types visible in the signature; prose that restates conditionals without explaining why they matter; prose that says an approach is better without explaining the mechanism, invariant, or constraint that makes it work.
5. Test organization: Tests after implementation, not before?
6. Proper noweb syntax: [[code]] notation in prose? Identifiers in chunk titles escaped with [[...]]? Valid chunk references?
7. Overview quality: For non-trivial files, does the overview identify the main parts, how they relate, and where the document will go next? Is the level appropriate for a maintainer who knows the language but not this codebase?
8. Visual clarity: When the structure is distributed, non-linear, or hard to track in prose, does the document add a roadmap or diagram?
9. Overview maintenance: When structure, flow, chunk organization, entry points, or likely edit locations changed, were the overview, local intro prose, roadmap figures, and any relevant README.md kept in sync?
10. Co-location and consistency: Are wiring chunks (attribute assignments, register/install calls, decorator-style mutation) placed next to the entities they mutate, not the helpers they call? When the same task is solved in multiple places in the file, do they share a pattern — or does prose explain why a particular site diverges?

Core Philosophy

Literate programming (Knuth) has two goals:

1. Explain to human beings what we want a computer to do
2. Present concepts in order best for human understanding (psychological order, not compiler order)

Variation Theory

Apply variation-theory skill when structuring explanations:

• Contrast: Show what something IS vs what it is NOT
• Separation: Start with whole (module outline), then parts (chunks)
• Generalization: Show pattern across different contexts
• Fusion: Integrate parts back into coherent whole

When the literate document is student-facing educational LaTeX, keep pedagogical meta-commentary such as variation/invariance labels and sequencing rationale out of the visible narrative. Put that reasoning in \ltnote{...} via the didactic-notes skill.

\ltnote{...} is also available in an ordinary (maintainer-facing) literate program, but scope it tightly: it is for reasoning about the exposition, not the code. A note on why the narrative is ordered as it is — why function A is explained before function B, why a concept is introduced in this chapter rather than later, why the whole precedes the parts — is commentary about the writing and belongs in an \ltnote. A design decision about the code itself (an architecture choice, why a data structure was picked, the rationale and any citations behind it) is content: it belongs in the visible body prose. The test: if removing the note would lose information about the program, it is body text; if it would only lose information about how the explanation was authored, it is an \ltnote. See the didactic-notes skill for details.

If the .nw document generates slides or handouts with live questions, Mentipy is a suitable tool for embedding those prompts in the LaTeX output.

CRITICAL: Show concrete examples FIRST, then state general principles. Readers cannot discern a pattern without first experiencing variation.

Noweb File Format

Documentation Chunks

• Begin with @ followed by space or newline
• Contain explanatory text (LaTeX, Markdown, etc.)
• Copied verbatim by noweave

Code Chunks

• Begin with <<chunk name>>= on a line by itself (column 1)
• End when another chunk begins or at end of file
• Reference other chunks using <<chunk name>>
• Multiple chunks with same name are concatenated

Syntax Rules

• Quote code in documentation using [[code]] (escapes LaTeX special chars). Never manually escape characters (e.g. \_) inside [[...]] — noweb handles all escaping automatically. Writing [[\_]] double-escapes.

• [[...]] works inside \item[...] labels (and other moving arguments), but separate the inner ]] from the outer ] with at least one character — typically a space. The failure mode is three brackets in a row (]]]), where noweb's ]] terminator and LaTeX's ] argument terminator collide and produce a Runaway argument? ! Paragraph ended before \@item was complete. error.

  GOOD — ]] is followed by a space or by other text:

  \item[The [[--restart]] flag] explains the idempotence rule.
  \item[ [[shell-basics $]] ] anchors on the nested shell prompt.
  

  BAD — [[...]] butts directly against the closing ]:

  \item[The required pattern [[shell-basics $]]]   % Runaway argument
  

  The same rule applies to any other LaTeX macro argument that is delimited with brackets (\caption[...], \section[...] short forms, etc.). When in doubt, add a trailing space inside the outer brackets.

• Escape: @<< for literal <<, @@ in column 1 for literal @

Don't Write @ Between Adjacent Code Chunks

@ starts a documentation chunk; it is not a code-chunk terminator. A code chunk already ends the moment any chunk begins — including another named code chunk. So when one code chunk is immediately followed by another with no prose between them, no @ is needed: the second chunk closes the first on its own.

Worse, a stray @ (followed by a blank line) opens an empty documentation chunk, which noweave renders as unwanted vertical space in the woven PDF.

Rule of thumb: write @ only when documentation prose follows the code. This is the complement of Writing Guideline 14 (prose-follows ⇒ @; code-follows ⇒ no @).

GOOD — the second chunk ends the first; no @, no empty doc chunk:

<<constants>>=
CONST = 3.14
<<functions>>=
def area_of_circle(radius):
    return CONST * radius ** 2
@

BAD — redundant @ opens an empty doc chunk → unwanted space in the PDF:

<<constants>>=
CONST = 3.14
@

<<functions>>=
def area_of_circle(radius):
    return CONST * radius ** 2
@

Embedding Heredocs — Keep the Delimiter Pair in the Wrapper Chunk

When a tangled script embeds a heredoc (a shell <<EOF ... EOF, an inline python3 - <<EOF, a cat > file <<EOF, etc.), put both the opening <<DELIM and the closing DELIM in the same wrapper chunk, with a single chunk reference for the body in between. Define the body in its own chunk that contains only the payload — never the delimiter.

This matters because noweb chunk boundaries and the heredoc are two independent layers: noweb blindly concatenates every definition of the body chunk, and the shell only sees the tangled text. If the closing delimiter is written inside the body chunk (especially when the body is built from several concatenated definitions), the heredoc terminates after the first piece and everything after it tangles as code instead of data — a syntax error far from its cause.

GOOD — delimiter pair together in the wrapper; body chunk holds only the payload (and may be split across several definitions safely):

<<write the config>>=
cat > "$file" <<'_EOF'
<<config body>>
_EOF
@

<<config body>>=
first line of payload
@

<<config body>>=
more payload, concatenated by noweb into the same heredoc
@

BAD — the closing delimiter lives in the body chunk, so a second definition (or any following chunk) escapes the heredoc:

<<write the config>>=
cat > "$file" <<'_EOF'
<<config body>>
@

<<config body>>=
first line of payload
_EOF
@

<<config body>>=
this never reaches the heredoc — it tangles as shell
@

Quote the opening delimiter (<<'_EOF') when the payload must not undergo shell expansion — for example embedded Python or JSON containing $. A distinctive delimiter such as _EOF or PYEOF keeps it easy to see the pair at a glance.

Writing Guidelines

1. Start with the human story - problem, approach, design decisions

2. Give the whole before the parts - substantial maintainer-facing documents should start with an overview of the moving pieces, their relationships, and the route through the file

3. Introduce concepts in pedagogical order - not compiler order

4. Use meaningful chunk names - 2-5 word summary of purpose (like pseudocode).

   Bucket chunks are named with nouns; phase chunks may be imperative. A bucket chunk holds the same kind of content across many definitions and is named as a category: <<constants>>, <<functions>>, <<imports>>, <<test functions>>, <<repo methods>>, <<option completions>>. Phase sub-chunks (Guideline 8) are different — they describe a step in an algorithm and read naturally as imperatives (<<pick random starting point>>, <<collect sentences until target length>>).

   BAD — verb-phrase bucket name:

   <<wire option completions>>=
   LAYOUT_OPTION.autocompletion = _choice_completer(LAYOUT_CHOICES)
   @
   

   GOOD — noun bucket name:

   <<option completions>>=
   LAYOUT_OPTION.autocompletion = _choice_completer(LAYOUT_CHOICES)
   @
   

5. Escape all identifiers in chunk names — any identifier (variable, function, parameter, attribute) that appears in a chunk title must be wrapped in [[...]]. This tells noweave to render it as code (monospace) and prevents LaTeX errors from underscores.

   BAD — bare identifier, underscore breaks LaTeX:

   <<restrict to top-level when not show_tree>>=
   if not show_tree:
       ...
   @
   

   GOOD — identifier escaped with [[...]]:

   <<restrict to top-level when not [[show_tree]]>>=
   if not show_tree:
       ...
   @
   

   Other examples: <<add graders to [[graders]] list>>, <<initialise [[default_username]]>>.

   IMPORTANT: Do not escape underscores or other special characters inside [[...]]. The brackets already tell noweb to escape everything. is correct; is wrong and will double-escape.

IMPORTANT: When a constant exists only to support one helper or one small cluster of related helpers, place its <<constants>>= bucket immediately adjacent to the bucket that contains those helpers. Do not hide configuration keys in one distant global block if the reader only needs them to understand one local section.

Preferred pattern — prose, then helper bucket, then local constants bucket for that helper family:

We read the debug flag from configuration rather than the process
environment so detached hooks and interactive commands see the same
value.

<<helper functions>>=
def track_debug_enabled():
    return config.get(TRACK_DEBUG_CONFIG)
<<constants>>=
TRACK_DEBUG_CONFIG = "track.debug"
@

This keeps the constant close enough to the narrative that readers meet it when they need it, without forcing them to search a giant global constants block.

BAD — all constants dumped in one subsection:

\subsection{Constants}

<<constants>>=
DATA_DIR = ...        % used in loading section
GUTENBERG_START = ... % used in extraction section
SENTENCE_RE = ...     % used in sentence splitting section
KEEP_PUNCT = ...      % used in normalization section
@

GOOD — each constant near the code that uses it:

\subsection{Loading Texts}

<<constants>>=
DATA_DIR = Path(__file__).parent / "data"
<<functions>>=
def load_text(path): ...
@

\subsection{Extracting Body Text}

<<constants>>=
GUTENBERG_START = "*** START OF"
GUTENBERG_END = "*** END OF"
<<functions>>=
def extract_body(text): ...
@

11. Co-locate wiring chunks with the entities they mutate, not the helpers they call. A wiring chunk (attribute assignment, register()/install() call, decorator-style mutation) has two natural neighbours: the helper functions it calls and the objects it mutates. Place it next to the objects it mutates so a reader who lands on the option/class/registry sees what is attached to it; the helper has its own section and does not benefit from a trailing assignment block. Chunk source order is independent of tangle order, so co-locating the source costs nothing at runtime.

BAD — wiring sits next to the helper it calls, far from the option it mutates:

<<constants>>=
LAYOUT_OPTION = typer.Option("slide", "--layout", help=LAYOUT_HELP)
@

...300 lines later...

<<functions>>=
def _choice_completer(values): ...
<<option completions>>=
LAYOUT_OPTION.autocompletion = _choice_completer(LAYOUT_CHOICES)
@

GOOD — wiring sits next to the object it mutates; the helper moves to live near the same neighbourhood:

<<constants>>=
LAYOUT_OPTION = typer.Option("slide", "--layout", help=LAYOUT_HELP)
<<functions>>=
def _choice_completer(values): ...
<<option completions>>=
LAYOUT_OPTION.autocompletion = _choice_completer(LAYOUT_CHOICES)
@

12. Define constants for magic numbers - never hardcode values

13. Co-locate dependencies with features - feature's imports in feature's section

14. Never leave prose inside an open code chunk — When inserting local <<constants>>= buckets or explanatory paragraphs, first close the current code chunk with @. Documentation between two function sections must be outside code mode; otherwise noweb tangles the prose into Python and produces syntax errors.

    BAD — prose accidentally tangled as Python:

    <<helper functions>>=
    def get_default_daily_limit():
        ...
    
    This helper uses the daily-limit config key.
    
    <<constants>>=
    DEFAULT_DAILY_LIMIT_CONFIG = "track.daily_limit"
    @
    

    GOOD — close the code chunk before prose, then reopen a bucket:

    <<helper functions>>=
    def get_default_daily_limit():
        ...
    @
    
    This helper uses the daily-limit config key.
    
    <<constants>>=
    DEFAULT_DAILY_LIMIT_CONFIG = "track.daily_limit"
    @
    

    Conversely, when a code chunk is immediately followed by another code chunk with no prose between them, omit the @ — see "Don't Write @ Between Adjacent Code Chunks" above.

15. Prefer public functions - Default to making functions public with docstrings. Only use _-prefixed private functions for true internal helpers tightly coupled to a single caller. Public utilities (e.g., normalize_text, letters_only) are reusable across modules and discoverable via help(). Duplicated private helpers across modules (e.g., _to_ascii in both vigenere.nw and plaintexts.nw) are a sign the function should be public in a shared module.

16. Keep lines under 80 characters - both prose and code

17. Survey siblings before introducing a new pattern. Before reaching for a new mechanism in a .nw file, grep the same file (and sibling .nw files) for places that already solve the same task. If three command-local options already attach completion via autocompletion=... at construction, the fourth option should match — or the prose must explain why the divergence is necessary (forward-reference constraint, chunk-tangle order, dependency on a later-defined symbol, etc.). The cost of this survey is one grep; the cost of skipping it is an asymmetric file where readers cannot tell whether the new mechanism is load-bearing or accidental.

LaTeX Documentation Quality

Apply latex-writing skill. Most common anti-patterns in .nw files:

Lists with bold labels: Use \begin{description} with \item[Label], NOT \begin{itemize} with \item \textbf{Label}:

Code with manual escaping: Use [[code]], NOT \texttt{...\_...}

Manual quotes: Use \enquote{...}, NOT "..." or ...'' ``

Manual cross-references: Use \cref{...}, NOT Section~\ref{...}

TikZ without preamble support: If a .nw file introduces a TikZ diagram, add \usepackage{tikz} to the project's preamble.tex.

Non-ASCII bytes in code chunks weave into the .tex: noweave copies code chunks verbatim, so any non-ASCII character inside a chunk — a box-drawing diagram in a test fixture, a Unicode normalization table, fancy quotes — lands in the woven LaTeX and must be typeset by the engine. How that behaves is engine-dependent: under pdfLaTeX an unmapped byte is a fatal ! LaTeX Error: Unicode character … (U+XXXX) not set up (at the woven line); under XeLaTeX/LuaLaTeX it is at worst a non-fatal Missing character warning. Check the build engine first — a literate project's build rules are often tangled from a .nw (e.g. tex.mk), so a make/submodule update can flip latexmk -pdf to -xelatex and change which fix is correct.

Fix it in the project's preamble.tex, not by deleting or escaping the byte in the .nw — the tangled program usually needs the real character at runtime (the test that asserts an ASCII-art figure is filtered, the table that maps real typographic punctuation). Before touching any non-ASCII byte in a code chunk, confirm whether the generated code depends on it. The preamble fix differs per engine (\DeclareUnicodeCharacter under pdfLaTeX — defined via \usepackage[utf8]{inputenc}; \newunicodechar under XeLaTeX/LuaLaTeX; guard with iftex). See the latex-writing skill's references/unicode-and-fonts.md for the full engine matrix, the T1-fontenc requirement for pdfLaTeX monospace fonts like Bera Mono, and the pdffonts/pdftotext diagnostics.

Progressive Disclosure Pattern

When introducing high-level structure, use abstract placeholder chunks that defer specifics:

def cli_show(user_regex,
             <<options for filtering>>):
  <<implementation>>
@

[... later, explain each option ...]

\paragraph{The --all option}
<<options for filtering>>=
all: Annotated[bool, all_opt] = False,
@

Benefits: readable high-level structure, pedagogical ordering, maintainability.

The same technique applies to function bodies: long functions can use <<phase name>> sub-chunks to present algorithmic steps in pedagogical order with prose between them (see Writing Guideline 8, "Decompose long functions").

Chunk Concatenation Patterns

Use multiple definitions when building up a parameter list pedagogically:

\subsection{Adding the diff flag}
<<args for diff>>=
diff=args.diff,
@

[... later ...]

\subsection{Fine-tuning thresholds}
<<args for diff>>=
threshold=args.threshold
@

Use separate chunks when contexts differ (different scopes):

<<args from command line>>=  # Has args object
diff=args.diff,
@

<<params for recursion>>=    # No args, only parameters
diff=diff,
@

Chunk Dependency Hazards

When a chunk references a variable defined in another chunk, there is an implicit dependency between them. Unlike function calls, noweb does not enforce that prerequisite chunks are included — the compiler only sees the tangled output. This makes it easy to reuse a chunk in a new code path while accidentally omitting the chunk that defines a variable it needs.

Rule: When reusing a chunk in a new code path, verify that all variables it references are defined by preceding chunks in that path.

BAD — chunk B depends on a variable set in chunk A, but a new path omits A:

<<path one>>=
<<chunk A>>
<<chunk B>>
@

<<path two>>=
<<chunk B>>      % ← UnboundLocalError: variable from A missing
@

GOOD — either include the prerequisite or document the dependency:

<<path two>>=
<<chunk A>>
<<chunk B>>
@

Tip: If a chunk both defines variables AND performs side effects that are not always wanted, consider splitting it so the variable-defining part can be included independently.

Design rule: Prefer extracting shared logic into a function rather than a reusable chunk when the logic is used across multiple code paths. Functions make dependencies explicit through parameters — a missing argument is a compile-time error, while a missing prerequisite chunk is only caught at runtime. Reserve chunks for pedagogical decomposition (presenting code in narrative order); use functions for operational decomposition (sharing logic between code paths).

Test Organization

CRITICAL: Tests MUST appear AFTER implementation, distributed throughout the file near the code they verify. NEVER create a \section{Tests} or \section{Unit Tests} that groups all tests at the end of the file.

See references/testing-patterns.md for detailed patterns.

Key rules:

• Each implementation section is followed by its <<test functions>>= chunk
• Use single <<test functions>> chunk name — noweb concatenates them
• If <<test functions>>= introduces a test class, treat it as the class shell and accumulate methods in a class-specific bucket chunk such as <<feature a test methods>>
• Use from module import * in the test file header
• Frame tests pedagogically: "Let's verify this works..."

BAD — all tests collected at the end:

\section{Encryption}
<<functions>>=
def encrypt(text, key): ...
@

\section{Decryption}
<<functions>>=
def decrypt(text, key): ...
@

\section{Tests}          % ← NEVER do this

<<test functions>>=
def test_encrypt(): ...
def test_decrypt(): ...
@

GOOD — each test immediately after its implementation:

\section{Encryption}
<<functions>>=
def encrypt(text, key): ...
@

Let's verify that encryption produces the expected ciphertext:

<<test functions>>=
def test_encrypt(): ...
@

\section{Decryption}
<<functions>>=
def decrypt(text, key): ...
@

We can verify that decryption inverts encryption:

<<test functions>>=
def test_decrypt(): ...
@

If a test section needs a test class, introduce the class in <<test functions>>= and collect its methods in a class-specific bucket chunk rather than concatenating indented methods directly into <<test functions>>=. See references/testing-patterns.md for a worked example and anti-pattern.

Multi-Directory Projects

For large projects (5+ .nw files), see references/multi-directory-projects.md.

Key structure:

project/
├── Makefile       # Root orchestrator (compile → test → docs)
├── pyproject.toml # Poetry packaging configuration
├── src/           # .nw files → .py + .tex
├── doc/           # Document wrapper (.nw), preamble.tex
├── tests/         # Extracted test files (unit/ subdir)
└── makefiles/     # Shared build rules (noweb.mk, subdir.mk)

Initializing a New Project

See references/project-initialization.md for full details. Quick checklist:

1. Create pyproject.toml with [tool.poetry] packages/include/exclude
2. Create src/.gitignore (*.py, *.tex) and tests/.gitignore (*.py)
3. Create src/packagename/Makefile with explicit __init__.py rule
4. Create src/packagename/packagename.nw with <<[[__init__.py]]>> and <<test [[packagename.py]]>> chunks
5. Create tests/Makefile with auto-discovery (uses %20 encoding, cpif, unit/ subdirectory)
6. Create doc/packagename.nw wrapper, doc/Makefile, doc/preamble.tex
7. Create root Makefile orchestrating compile → test → docs

LaTeX-Safe Chunk Names

Use [[...]] notation for Python chunks with underscores:

<<[[module_name.py]]>>=
def my_function():
    pass
@

Extract with: notangle -R"[[module_name.py]]" file.nw > module_name.py

Filename Chunk Names Drive Language Inference

Naming root chunks exactly like their output filenames is load-bearing, not just a convention: the autolang and tominted filters infer each chunk's language from filename-like names — Python for <<[[fib.py]]>>, Make for <<[[Makefile]]>> — and chunks whose names are not filenames (<<functions>>, <<test functions>>) inherit the language of the chunks that use them. This is what lets one .nw file mix languages (a program plus its Makefile) while each chunk is highlighted by its own lexer and the identifier index stays free of foreign identifiers. See references/noweb-commands.md for the inference rules and .ext=lexer / name=lexer mapping overrides.

Best Practices Summary

1. Write documentation first - then add code

2. Keep lines under 80 characters

3. Check for unused chunks - run noroots to find typos

4. Keep tangled code in .gitignore - .nw is source of truth

5. NEVER commit generated files - .py and .tex from .nw are build artifacts

6. Test your tangles - ensure extracted code runs

7. Require PEP-257 docstrings on all public functions - Prose in .nw is for maintainers reading the literate source; docstrings are for users of the compiled .py who never see the .nw file. Both are needed. Private functions (prefixed _) may omit docstrings. Never use \cref or other LaTeX commands inside docstrings — and never use noweb [[...]] code-quoting there either. [[...]] is only interpreted by noweave in documentation chunks; inside a docstring (a string in a code chunk) it is copied verbatim, so it leaks into the generated .py and help() output. Quote code in docstrings however your project's docstrings already do (plain text, markdown `x`, or RST x ) — just not with [[...]].

   Chunk references (<<...>>) inside docstrings are fine: the standard weave's custom lexer (tominted -lexer noweb_lexer.py) keeps them hyperlinked in the woven PDF even though they sit inside a Python string literal.

   BAD — noweb [[...]] in a docstring leaks into help():

   <<functions>>=
   def __setattr__(self, name, value):
       """Keep [[labels]] a sorted, duplicate-free list."""
       ...
   @
   

   GOOD — quote code with the project's docstring convention:

   <<functions>>=
   def __setattr__(self, name, value):
       """Keep `labels` a sorted, duplicate-free list."""
       ...
   @
   

   BAD — function with prose but no docstring:

   We convert text to lowercase ASCII for uniform comparison.
   
   <<functions>>=
   def normalize_text(text):
       return text.lower().encode("ascii", "ignore").decode()
   @
   

Git Workflow

See references/git-workflow.md for details.

Core rules:

• Only commit .nw files to git
• Add generated files to .gitignore immediately
• Regenerate code with make after checkout/pull
• Never commit generated .py or .tex files

Noweb Commands Quick Reference

See references/noweb-commands.md for details, including the one-time custom-lexer setup the standard weave below depends on.

# Tangling
notangle -R"[[module.py]]" file.nw > module.py
notangle -t8 -R"[[Makefile]]" file.nw > Makefile  # -t8 keeps tabs
noroots file.nw                              # List root chunks

# Weaving (standard recipe: highlighted with minted, clean index)
# For inclusion; the master preamble loads minted via \usepackage[minted]{noweb}
noweave -n -delay -autolang -autodefs python3 -autodefs sh \
    -autodefs make -index \
    -filter 'tominted -lexer noweb_lexer.py' file.nw > file.tex
# compile with -shell-escape (pdflatex -shell-escape or latexmk option)
# Standalone (noweave writes the preamble): add -minted so it loads minted
noweave -autolang -autodefs python3 -autodefs sh -autodefs make \
    -index -minted file.nw > file.tex

# Classic fallback (no minted/patched noweb; identifier uses inside
# code stay hyperlinked, at the price of no syntax highlighting)
noweave -n -delay -x -t2 file.nw > file.tex  # For inclusion
noweave -latex -x file.nw > file.tex         # Standalone

Pipeline rules baked into the standard recipe: -autolang always runs first no matter where it appears on the command line; tominted must be the last filter, after -index. tominted is LaTeX-only — never combine it with -html.

Stacking several -autodefs works because -autolang's @language annotations gate each filter to its own chunks. Available filters: python3, sh/bash, make, haskell, rust, java, c (ANSI C chunks only), cpp (C++ chunks: classes, templates, namespaces, qualified methods), plus the classic icon/sml/tex/yacc/etc. — run noweave -showautodefs for the installed list. Filter budget: the dbosk fork's pipeline is unbounded — it accumulates filters in a single variable and joins them with |, so any number of -autodefs filters can stack alongside -autolang, -index (which inserts two filters, finduses and noidx) and tominted in one weave. A six-language document is one pass, not two. (Stock and pre-fork noweave enumerated only seven filter slots, a portability artifact of the array-less /bin/sh pipeline builder; that ceiling — three -autodefs for a highlighted indexed weave, four for a classic one — applies only if you run against an older noweave, in which case split the languages across a second weave; see the multilang example in noweb's examples/.)

[[code]] in headings + hyperref (the \hyper@@link / \Hy@reserved@a error)

A [[identifier]] quote inside a \chapter/\section/\subsection heading is a hazard when -index is active. -index turns the quote into a hyperlinked identifier use (\nwlinkedidentq → \nwhyperreference → \hyperlink). In a heading the title is expanded into moving arguments — the .toc, the running header, and the PDF bookmark — and hyperref breaks there. Under newer TeX Live (≥2025) the symptom is a fatal error, typically:

! Undefined control sequence.
\hyper@@link ->\let \Hy@reserved@a ...
l.379 ...}}{NW4Ar8dO-3qo3Hb-8}\nwendquote} method}

(preceded by a cascade of Token not allowed in a PDF string ... removing '\Hy@reserved@a' warnings), or a stack overflow from noweb's self-redefining \nwhyperreference (its terminating \global\let only fires when the macro is executed, never when it is merely expanded into a moving argument). This is a toolchain regression: the same document built cleanly under older hyperref/TeX Live, so a make/submodule update that bumps TeX Live can surface it with no source change. Only headings whose quoted text is an indexed identifier (a name defined in a code chunk, e.g. a dunder like __str__) become links and fail; plain words like [[command]] weave to a non-linked \Tt and are harmless.

Preferred fix — keep the literate [[...]]-in-headings convention by making the link robust project-wide in preamble.tex (after hyperref is loaded). This is two cooperating fixes; both are needed, and they fix every current and future heading at once instead of editing each .nw:

% (1) robust \nwhyperreference: written literally into moving arguments instead
%     of expanding (and misfiring its self-\let) there, yet still links in body
\DeclareRobustCommand\nwhyperreference[2]{\hyperlink{noweb.#1}{#2}}
% (2) strip noweb's code-quote macros from PDF bookmark strings → clean bookmarks
\pdfstringdefDisableCommands{%
  \def\nwlinkedidentq#1#2{#1}%
  \let\Tt\relax
  \let\nwendquote\relax
}

This block is in the shipped references/preamble.tex. The narrower alternative — rewriting the heading to use \texttt{...} instead of [[...]] — also works but loses the cross-reference link and must be repeated per heading, so prefer the preamble fix.

While debugging such a build, note latexmk can mask the real LaTeX error behind a later toolchain failure (e.g. biber: malformed bcf from a stale aux). Run one bare pdflatex -halt-on-error pass to see the first genuine error before chasing bibliography/pythontex noise.

When Literate Programming Is Valuable

• Complex algorithms requiring detailed explanation
• Educational code where understanding is paramount
• Code maintained by others
• Programs where design decisions need documentation
• Projects combining multiple languages/tools