recfold 1.0.0__tar.gz

This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.
@@ -0,0 +1,220 @@
1
+ # Byte-compiled / optimized / DLL files
2
+ __pycache__/
3
+ *.py[codz]
4
+ *$py.class
5
+
6
+ # C extensions
7
+ *.so
8
+
9
+ # Distribution / packaging
10
+ .Python
11
+ build/
12
+ develop-eggs/
13
+ dist/
14
+ downloads/
15
+ eggs/
16
+ .eggs/
17
+ lib/
18
+ lib64/
19
+ parts/
20
+ sdist/
21
+ var/
22
+ wheels/
23
+ share/python-wheels/
24
+ *.egg-info/
25
+ .installed.cfg
26
+ *.egg
27
+ MANIFEST
28
+
29
+ # PyInstaller
30
+ # Usually these files are written by a python script from a template
31
+ # before PyInstaller builds the exe, so as to inject date/other infos into it.
32
+ *.manifest
33
+ *.spec
34
+
35
+ # Installer logs
36
+ pip-log.txt
37
+ pip-delete-this-directory.txt
38
+
39
+ # Unit test / coverage reports
40
+ htmlcov/
41
+ .tox/
42
+ .nox/
43
+ .coverage
44
+ .coverage.*
45
+ .cache
46
+ nosetests.xml
47
+ coverage.xml
48
+ *.cover
49
+ *.py.cover
50
+ *.lcov
51
+ .hypothesis/
52
+ .pytest_cache/
53
+ cover/
54
+
55
+ # Translations
56
+ *.mo
57
+ *.pot
58
+
59
+ # Django stuff:
60
+ *.log
61
+ local_settings.py
62
+ db.sqlite3
63
+ db.sqlite3-journal
64
+
65
+ # Flask stuff:
66
+ instance/
67
+ .webassets-cache
68
+
69
+ # Scrapy stuff:
70
+ .scrapy
71
+
72
+ # Sphinx documentation
73
+ docs/_build/
74
+
75
+ # PyBuilder
76
+ .pybuilder/
77
+ target/
78
+
79
+ # Jupyter Notebook
80
+ .ipynb_checkpoints
81
+
82
+ # IPython
83
+ profile_default/
84
+ ipython_config.py
85
+
86
+ # pyenv
87
+ # For a library or package, you might want to ignore these files since the code is
88
+ # intended to run in multiple environments; otherwise, check them in:
89
+ # .python-version
90
+
91
+ # pipenv
92
+ # According to pypa/pipenv#598, it is recommended to include Pipfile.lock in version control.
93
+ # However, in case of collaboration, if having platform-specific dependencies or dependencies
94
+ # having no cross-platform support, pipenv may install dependencies that don't work, or not
95
+ # install all needed dependencies.
96
+ # Pipfile.lock
97
+
98
+ # uv
99
+ # Similar to Pipfile.lock, it is generally recommended to include uv.lock in version control.
100
+ # This is especially recommended for binary packages to ensure reproducibility, and is more
101
+ # commonly ignored for libraries.
102
+ # uv.lock
103
+
104
+ # poetry
105
+ # Similar to Pipfile.lock, it is generally recommended to include poetry.lock in version control.
106
+ # This is especially recommended for binary packages to ensure reproducibility, and is more
107
+ # commonly ignored for libraries.
108
+ # https://python-poetry.org/docs/basic-usage/#commit-your-poetrylock-file-to-version-control
109
+ # poetry.lock
110
+ # poetry.toml
111
+
112
+ # pdm
113
+ # Similar to Pipfile.lock, it is generally recommended to include pdm.lock in version control.
114
+ # pdm recommends including project-wide configuration in pdm.toml, but excluding .pdm-python.
115
+ # https://pdm-project.org/en/latest/usage/project/#working-with-version-control
116
+ # pdm.lock
117
+ # pdm.toml
118
+ .pdm-python
119
+ .pdm-build/
120
+
121
+ # pixi
122
+ # Similar to Pipfile.lock, it is generally recommended to include pixi.lock in version control.
123
+ # pixi.lock
124
+ # Pixi creates a virtual environment in the .pixi directory, just like venv module creates one
125
+ # in the .venv directory. It is recommended not to include this directory in version control.
126
+ .pixi/*
127
+ !.pixi/config.toml
128
+
129
+ # PEP 582; used by e.g. github.com/David-OConnor/pyflow and github.com/pdm-project/pdm
130
+ __pypackages__/
131
+
132
+ # Celery stuff
133
+ celerybeat-schedule*
134
+ celerybeat.pid
135
+
136
+ # Redis
137
+ *.rdb
138
+ *.aof
139
+ *.pid
140
+
141
+ # RabbitMQ
142
+ mnesia/
143
+ rabbitmq/
144
+ rabbitmq-data/
145
+
146
+ # ActiveMQ
147
+ activemq-data/
148
+
149
+ # SageMath parsed files
150
+ *.sage.py
151
+
152
+ # Environments
153
+ .env
154
+ .envrc
155
+ .venv
156
+ env/
157
+ venv/
158
+ ENV/
159
+ env.bak/
160
+ venv.bak/
161
+
162
+ # Spyder project settings
163
+ .spyderproject
164
+ .spyproject
165
+
166
+ # Rope project settings
167
+ .ropeproject
168
+
169
+ # mkdocs/Zensical documentation
170
+ /site
171
+
172
+ # mypy
173
+ .mypy_cache/
174
+ .dmypy.json
175
+ dmypy.json
176
+
177
+ # Pyre type checker
178
+ .pyre/
179
+
180
+ # pytype static type analyzer
181
+ .pytype/
182
+
183
+ # Cython debug symbols
184
+ cython_debug/
185
+
186
+ # PyCharm
187
+ # JetBrains specific template is maintained in a separate JetBrains.gitignore that can
188
+ # be found at https://github.com/github/gitignore/blob/main/Global/JetBrains.gitignore
189
+ # and can be added to the global gitignore or merged into this file. For a more nuclear
190
+ # option (not recommended) you can uncomment the following to ignore the entire idea folder.
191
+ # .idea/
192
+
193
+ # Abstra
194
+ # Abstra is an AI-powered process automation framework.
195
+ # Ignore directories containing user credentials, local state, and settings.
196
+ # Learn more at https://abstra.io/docs
197
+ .abstra/
198
+
199
+ # Visual Studio Code
200
+ # Visual Studio Code specific template is maintained in a separate VisualStudioCode.gitignore that
201
+ # can be found at https://github.com/github/gitignore/blob/main/Global/VisualStudioCode.gitignore
202
+ # and can be added to the global gitignore or merged into this file. However, if you prefer, you
203
+ # could uncomment the following to ignore the entire vscode folder
204
+ # .vscode/
205
+ # Temporary file for partial code execution
206
+ tempCodeRunnerFile.py
207
+
208
+ # Ruff stuff:
209
+ .ruff_cache/
210
+
211
+ # PyPI configuration file
212
+ .pypirc
213
+
214
+ # Marimo
215
+ marimo/_static/
216
+ marimo/_lsp/
217
+ __marimo__/
218
+
219
+ # Streamlit
220
+ .streamlit/secrets.toml
recfold-1.0.0/LICENSE ADDED
@@ -0,0 +1,21 @@
1
+ MIT License
2
+
3
+ Copyright (c) 2026 Zenith-20
4
+
5
+ Permission is hereby granted, free of charge, to any person obtaining a copy
6
+ of this software and associated documentation files (the "Software"), to deal
7
+ in the Software without restriction, including without limitation the rights
8
+ to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
9
+ copies of the Software, and to permit persons to whom the Software is
10
+ furnished to do so, subject to the following conditions:
11
+
12
+ The above copyright notice and this permission notice shall be included in all
13
+ copies or substantial portions of the Software.
14
+
15
+ THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
16
+ IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
17
+ FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
18
+ AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
19
+ LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
20
+ OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
21
+ SOFTWARE.
recfold-1.0.0/PKG-INFO ADDED
@@ -0,0 +1,173 @@
1
+ Metadata-Version: 2.5
2
+ Name: recfold
3
+ Version: 1.0.0
4
+ Summary: Zero-runtime-dependency recursion call-tree analyzer and state inspector for Python.
5
+ Project-URL: Homepage, https://github.com/Zenith-20/recfold
6
+ Project-URL: Repository, https://github.com/Zenith-20/recfold
7
+ Author: Zenith-20, ssartayev, Alishnis, bekonty, exehamza, thecocogoat
8
+ License-Expression: MIT
9
+ License-File: LICENSE
10
+ Requires-Python: >=3.9
11
+ Description-Content-Type: text/markdown
12
+
13
+ # recfold
14
+
15
+ `recfold` is a Python package for inspecting recursive function calls in Jupyter notebooks or standalone HTML files. It records traced calls in a call tree, folds structurally matching subtrees into a directed acyclic graph (DAG), and lets you inspect arguments, return values, errors, and boundary snapshots for individual calls.
16
+
17
+ ## Scope and design choices
18
+
19
+ ### No third-party runtime dependencies
20
+
21
+ `recfold` uses only the Python standard library at runtime. Its visualization is generated as self-contained HTML, SVG, CSS, and JavaScript. Runtime use does not require Graphviz, native C binaries, external CDNs, browser extensions, or `ipywidgets`. Notebook display still depends on the frontend permitting sandboxed `srcdoc` iframes and inline scripts.
22
+
23
+ ### Tree-to-DAG folding and redundancy metrics
24
+
25
+ Tracing records calls that pass the configured depth and repeated-active-call guards. `recfold` then folds subtrees with matching structural keys into shared DAG nodes and calculates a structural redundancy ratio:
26
+
27
+ ```text
28
+ redundancy ratio = (recorded calls - unique folded subtrees) / recorded calls
29
+ ```
30
+
31
+ For `fib(5)`, `recfold` records 15 calls and folds them into 6 unique subtrees. The ratio is `(15 - 6) / 15 = 0.6`, or 60%.
32
+
33
+ This is a post-hoc structural summary of the recorded trace, not an optimization. Folding does not memoize the traced function, speed up execution, or skip function calls. Recursive calls still execute, and the raw call tree remains in memory.
34
+
35
+ ### Entry and exit snapshots
36
+
37
+ At call entry, `recfold` deep-copies positional and keyword arguments; at normal return, it deep-copies the result. If the function raises, the exit snapshot records the exception text. This can preserve a mutable argument, such as a backtracking path, as it appeared at each call boundary rather than exposing later mutations in earlier snapshots.
38
+
39
+ These are boundary snapshots, not a step-by-step debugger. The tracer does not automatically capture local variables or intermediate mutations. Each argument and result is copied separately; references shared across separately copied values are not preserved. Values that cannot be deep-copied are replaced with placeholder text.
40
+
41
+ ### Structural views with occurrence records
42
+
43
+ The renderer provides interactive **Call tree** and **Folded DAG** views. Select a node and use the **Recorded occurrence** dropdown to inspect the corresponding call ID, depth, parent ID, arguments, result or error, and entry/exit snapshots. The occurrences retain the original call records; call IDs are assigned in call-entry order.
44
+
45
+ ## Installation
46
+
47
+ Python 3.9 or newer is required. Install from a local clone:
48
+
49
+ ```sh
50
+ git clone https://github.com/Zenith-20/recfold.git
51
+ cd recfold
52
+ python -m pip install -e .
53
+ ```
54
+
55
+ If you are using JupyterHub, ensure the repository is on the server's filesystem and install it into your active kernel:
56
+
57
+ ```python
58
+ %pip install -e /path/on/server/to/recfold
59
+ ```
60
+
61
+ ## Quick start
62
+
63
+ Run this in a Jupyter notebook cell:
64
+
65
+ ```python
66
+ from recfold.tracer import trace
67
+ from recfold.renderer import render_session
68
+
69
+ @trace
70
+ def fib(n):
71
+ if n < 2:
72
+ return n
73
+ return fib(n - 1) + fib(n - 2)
74
+
75
+ answer = fib(5)
76
+ print(answer) # 5
77
+
78
+ widget = render_session(fib.last_session, title="Fibonacci: repeated work")
79
+ widget # Keep as the last expression so Jupyter displays the widget
80
+ ```
81
+
82
+ For a plain recursive function, `run_traced` returns the result and trace
83
+ session:
84
+
85
+ ```python
86
+ from recfold.tracer import run_traced
87
+ from recfold.renderer import render_session
88
+
89
+ def factorial(n):
90
+ if n < 2:
91
+ return 1
92
+ return n * factorial(n - 1)
93
+
94
+ answer, session = run_traced(factorial, 5)
95
+ widget = render_session(session)
96
+ widget
97
+ ```
98
+
99
+ ## Offline HTML export
100
+
101
+ To save the visualization as a standalone HTML document:
102
+
103
+ ```python
104
+ widget.save("fib.html")
105
+
106
+ # Use overwrite=True to replace an existing file
107
+ widget.save("fib.html", overwrite=True)
108
+ ```
109
+
110
+ The exported document contains the graph views, styles, scripts, and rendered trace data. It makes no network requests and can be opened offline in a JavaScript-enabled browser. Review the included trace data before sharing.
111
+
112
+ ## Current limitations
113
+
114
+ * **Synchronous execution only:** The decorator does not trace the execution of async function bodies or generator bodies. They return coroutine or generator objects when called; tracing does not await or iterate them.
115
+ * **Representation-based cycle guard:** The active-call guard compares function names and argument `repr()` strings, using the default 80-character limit. Distinct values with the same representation can be mistaken for a repeated active call. For terminating recursion where this is a concern, `@trace(detect_cycles=False, max_depth=...)` disables that guard while retaining a depth bound.
116
+ * **No automatic local-variable capture:** The tracer snapshots arguments at entry and return values or exceptions at exit. `snapshot_locals()` is available in `recfold.snapshot`, but the tracer does not call it.
117
+ * **Structural matching is not semantic proof:** Folding keys use function names, argument and result representations, exception text, and ordered child subtree keys. Matching keys do not establish mathematical equivalence or purity.
118
+ * **No virtualization:** Both graph views are rendered with all included nodes and edges. Large traces can require substantial memory and browser rendering work.
119
+
120
+ ## Project structure
121
+
122
+ ```text
123
+ models.py Dataclasses for calls, DAG nodes/edges, sessions, and statistics
124
+ snapshot.py Deep-copying and string truncation utilities
125
+ tracer.py @trace decorator, call stack tracking, and depth/cycle guards
126
+ analyzer.py Subtree folding logic and redundancy statistics
127
+ layout.py Pure-Python 2D node positioning and straight edge geometry
128
+ renderer.py Self-contained HTML/SVG generator and Jupyter widget wrapper
129
+ ```
130
+
131
+ There is no console-script entry point or package-level re-export; import
132
+ functions from their modules as shown above.
133
+
134
+ ## Development and testing
135
+
136
+ To run the unit test suite:
137
+
138
+ ```sh
139
+ python -m pip install pytest
140
+ python -m pytest
141
+ ```
142
+
143
+ ### Optional browser tests
144
+
145
+ End-to-end UI tests (verifying click selection, panning, zooming, and keyboard navigation in Chromium) require Playwright:
146
+
147
+ ```sh
148
+ python -m pip install playwright
149
+ python -m playwright install chromium
150
+ ```
151
+
152
+ Run them by setting the opt-in environment variable:
153
+
154
+ * **macOS / Linux:**
155
+ ```sh
156
+ RECFOLD_BROWSER_TESTS=1 python -m pytest tests/test_renderer.py
157
+ ```
158
+ * **Windows PowerShell:**
159
+ ```powershell
160
+ $env:RECFOLD_BROWSER_TESTS = "1"
161
+ python -m pytest tests/test_renderer.py
162
+ ```
163
+
164
+ Without this flag, browser tests are skipped automatically.
165
+
166
+ ## Further documentation
167
+
168
+ * [Renderer API, integration, and troubleshooting](docs/renderer.md)
169
+ * [Demo notebook: Fibonacci, backtracking, and error guards](docs/recfold_demo.ipynb)
170
+
171
+ ## License
172
+
173
+ [MIT](LICENSE)
@@ -0,0 +1,161 @@
1
+ # recfold
2
+
3
+ `recfold` is a Python package for inspecting recursive function calls in Jupyter notebooks or standalone HTML files. It records traced calls in a call tree, folds structurally matching subtrees into a directed acyclic graph (DAG), and lets you inspect arguments, return values, errors, and boundary snapshots for individual calls.
4
+
5
+ ## Scope and design choices
6
+
7
+ ### No third-party runtime dependencies
8
+
9
+ `recfold` uses only the Python standard library at runtime. Its visualization is generated as self-contained HTML, SVG, CSS, and JavaScript. Runtime use does not require Graphviz, native C binaries, external CDNs, browser extensions, or `ipywidgets`. Notebook display still depends on the frontend permitting sandboxed `srcdoc` iframes and inline scripts.
10
+
11
+ ### Tree-to-DAG folding and redundancy metrics
12
+
13
+ Tracing records calls that pass the configured depth and repeated-active-call guards. `recfold` then folds subtrees with matching structural keys into shared DAG nodes and calculates a structural redundancy ratio:
14
+
15
+ ```text
16
+ redundancy ratio = (recorded calls - unique folded subtrees) / recorded calls
17
+ ```
18
+
19
+ For `fib(5)`, `recfold` records 15 calls and folds them into 6 unique subtrees. The ratio is `(15 - 6) / 15 = 0.6`, or 60%.
20
+
21
+ This is a post-hoc structural summary of the recorded trace, not an optimization. Folding does not memoize the traced function, speed up execution, or skip function calls. Recursive calls still execute, and the raw call tree remains in memory.
22
+
23
+ ### Entry and exit snapshots
24
+
25
+ At call entry, `recfold` deep-copies positional and keyword arguments; at normal return, it deep-copies the result. If the function raises, the exit snapshot records the exception text. This can preserve a mutable argument, such as a backtracking path, as it appeared at each call boundary rather than exposing later mutations in earlier snapshots.
26
+
27
+ These are boundary snapshots, not a step-by-step debugger. The tracer does not automatically capture local variables or intermediate mutations. Each argument and result is copied separately; references shared across separately copied values are not preserved. Values that cannot be deep-copied are replaced with placeholder text.
28
+
29
+ ### Structural views with occurrence records
30
+
31
+ The renderer provides interactive **Call tree** and **Folded DAG** views. Select a node and use the **Recorded occurrence** dropdown to inspect the corresponding call ID, depth, parent ID, arguments, result or error, and entry/exit snapshots. The occurrences retain the original call records; call IDs are assigned in call-entry order.
32
+
33
+ ## Installation
34
+
35
+ Python 3.9 or newer is required. Install from a local clone:
36
+
37
+ ```sh
38
+ git clone https://github.com/Zenith-20/recfold.git
39
+ cd recfold
40
+ python -m pip install -e .
41
+ ```
42
+
43
+ If you are using JupyterHub, ensure the repository is on the server's filesystem and install it into your active kernel:
44
+
45
+ ```python
46
+ %pip install -e /path/on/server/to/recfold
47
+ ```
48
+
49
+ ## Quick start
50
+
51
+ Run this in a Jupyter notebook cell:
52
+
53
+ ```python
54
+ from recfold.tracer import trace
55
+ from recfold.renderer import render_session
56
+
57
+ @trace
58
+ def fib(n):
59
+ if n < 2:
60
+ return n
61
+ return fib(n - 1) + fib(n - 2)
62
+
63
+ answer = fib(5)
64
+ print(answer) # 5
65
+
66
+ widget = render_session(fib.last_session, title="Fibonacci: repeated work")
67
+ widget # Keep as the last expression so Jupyter displays the widget
68
+ ```
69
+
70
+ For a plain recursive function, `run_traced` returns the result and trace
71
+ session:
72
+
73
+ ```python
74
+ from recfold.tracer import run_traced
75
+ from recfold.renderer import render_session
76
+
77
+ def factorial(n):
78
+ if n < 2:
79
+ return 1
80
+ return n * factorial(n - 1)
81
+
82
+ answer, session = run_traced(factorial, 5)
83
+ widget = render_session(session)
84
+ widget
85
+ ```
86
+
87
+ ## Offline HTML export
88
+
89
+ To save the visualization as a standalone HTML document:
90
+
91
+ ```python
92
+ widget.save("fib.html")
93
+
94
+ # Use overwrite=True to replace an existing file
95
+ widget.save("fib.html", overwrite=True)
96
+ ```
97
+
98
+ The exported document contains the graph views, styles, scripts, and rendered trace data. It makes no network requests and can be opened offline in a JavaScript-enabled browser. Review the included trace data before sharing.
99
+
100
+ ## Current limitations
101
+
102
+ * **Synchronous execution only:** The decorator does not trace the execution of async function bodies or generator bodies. They return coroutine or generator objects when called; tracing does not await or iterate them.
103
+ * **Representation-based cycle guard:** The active-call guard compares function names and argument `repr()` strings, using the default 80-character limit. Distinct values with the same representation can be mistaken for a repeated active call. For terminating recursion where this is a concern, `@trace(detect_cycles=False, max_depth=...)` disables that guard while retaining a depth bound.
104
+ * **No automatic local-variable capture:** The tracer snapshots arguments at entry and return values or exceptions at exit. `snapshot_locals()` is available in `recfold.snapshot`, but the tracer does not call it.
105
+ * **Structural matching is not semantic proof:** Folding keys use function names, argument and result representations, exception text, and ordered child subtree keys. Matching keys do not establish mathematical equivalence or purity.
106
+ * **No virtualization:** Both graph views are rendered with all included nodes and edges. Large traces can require substantial memory and browser rendering work.
107
+
108
+ ## Project structure
109
+
110
+ ```text
111
+ models.py Dataclasses for calls, DAG nodes/edges, sessions, and statistics
112
+ snapshot.py Deep-copying and string truncation utilities
113
+ tracer.py @trace decorator, call stack tracking, and depth/cycle guards
114
+ analyzer.py Subtree folding logic and redundancy statistics
115
+ layout.py Pure-Python 2D node positioning and straight edge geometry
116
+ renderer.py Self-contained HTML/SVG generator and Jupyter widget wrapper
117
+ ```
118
+
119
+ There is no console-script entry point or package-level re-export; import
120
+ functions from their modules as shown above.
121
+
122
+ ## Development and testing
123
+
124
+ To run the unit test suite:
125
+
126
+ ```sh
127
+ python -m pip install pytest
128
+ python -m pytest
129
+ ```
130
+
131
+ ### Optional browser tests
132
+
133
+ End-to-end UI tests (verifying click selection, panning, zooming, and keyboard navigation in Chromium) require Playwright:
134
+
135
+ ```sh
136
+ python -m pip install playwright
137
+ python -m playwright install chromium
138
+ ```
139
+
140
+ Run them by setting the opt-in environment variable:
141
+
142
+ * **macOS / Linux:**
143
+ ```sh
144
+ RECFOLD_BROWSER_TESTS=1 python -m pytest tests/test_renderer.py
145
+ ```
146
+ * **Windows PowerShell:**
147
+ ```powershell
148
+ $env:RECFOLD_BROWSER_TESTS = "1"
149
+ python -m pytest tests/test_renderer.py
150
+ ```
151
+
152
+ Without this flag, browser tests are skipped automatically.
153
+
154
+ ## Further documentation
155
+
156
+ * [Renderer API, integration, and troubleshooting](docs/renderer.md)
157
+ * [Demo notebook: Fibonacci, backtracking, and error guards](docs/recfold_demo.ipynb)
158
+
159
+ ## License
160
+
161
+ [MIT](LICENSE)
@@ -0,0 +1,94 @@
1
+ # Member 6: renderer and documentation
2
+
3
+ ## Scope
4
+
5
+ Implemented the team's Member 6 brief in `src/recfold/renderer.py`, with tests in
6
+ `tests/test_renderer.py` and documentation in `README.md` and `docs/`.
7
+ No changes were made to the other members' modules, package exports, packaging
8
+ configuration, or release version. PyPI publication remains Member 1's task.
9
+
10
+ ## Function-by-function implementation
11
+
12
+ Each implementation commit was preceded by new failing tests, then a passing
13
+ test run, followed by a commit and push. Tests and their function were committed
14
+ together. The commits are listed in implementation order:
15
+
16
+ | Commit | Function / increment |
17
+ | --- | --- |
18
+ | `5e9656e` | `build_render_data`: raw-tree validation, DAG/tree payloads, occurrence snapshots, supplied layout support |
19
+ | `37e4c8f` | `render_svg`: accessible SVG nodes, arrows, labels, escaping, errors, and unique marker IDs |
20
+ | `ae5d980` | `render_html`: offline interactive views, inspector, keyboard controls, zoom/pan, responsive styling |
21
+ | `f022b03` | `RecursionWidget._repr_html_`: isolated Jupyter iframe display |
22
+ | `91bd5e0` | `RecursionWidget.save`: standalone HTML export with explicit overwrite |
23
+ | `ef1dc2a` | `render_session`: notebook-friendly public entry point |
24
+
25
+ Follow-up commits add boundary/browser coverage and documentation. Those are
26
+ validation/documentation work, not bundled implementation of additional functions.
27
+
28
+ ## Deliverables
29
+
30
+ - Toggle between original call tree and folded DAG.
31
+ - Select nodes by click or keyboard and highlight their incident edges.
32
+ - Inspect arguments, results, errors, parent IDs, and saved snapshots.
33
+ - Select individual occurrences of folded calls.
34
+ - View call counts, unique subtree counts, depth, and repeated-call percentage.
35
+ - Zoom, fit, and pan with buttons, pointer gestures, or keyboard shortcuts.
36
+ - Render multiple independent widgets in one notebook.
37
+ - Export a complete interactive HTML file that works offline.
38
+ - Use a precomputed layout without changing the supplied trace.
39
+ - Handle empty and partial/error traces, escape trace text, and reject malformed
40
+ raw trees before folding.
41
+
42
+ ## Validation on 2026-11-16
43
+
44
+ - **137 tests passed** with the optional browser checks enabled: the original
45
+ 98 tests plus 39 renderer test cases.
46
+ - The ordinary pytest run does not require Playwright; five browser tests are
47
+ explicitly opt-in through `RECFOLD_BROWSER_TESTS=1`.
48
+ - Chromium checks covered graph switching, occurrence selection, keyboard
49
+ selection, zoom/fit, pointer panning, mobile width, hostile trace text,
50
+ independent sandboxed widgets, empty traces, error details, and disclosures.
51
+ - Desktop and mobile screenshots were visually inspected.
52
+ - `docs/recfold_demo.ipynb` was validated and executed in a real Jupyter kernel:
53
+ all cells completed and four rich HTML widgets were produced. The committed
54
+ notebook is clean, with no saved outputs.
55
+ - In local **JupyterLab 4.6.6**, the notebook iframe loaded and its tree/DAG and
56
+ occurrence-selection controls worked. This does not establish compatibility
57
+ with every managed JupyterHub content-security policy.
58
+ - The wheel and source archive built successfully. A wheel smoke test under
59
+ Python's `-S` mode imported and rendered a trace without site-packages; wheel
60
+ metadata declared no runtime dependencies.
61
+ - Runtime validation used **Python 3.12**. The renderer also passed Python 3.9
62
+ syntax parsing; an actual Python 3.9 runtime was not tested.
63
+ - CodeRabbit review was unavailable because the local CLI was signed out;
64
+ manual review and the automated checks above were used instead.
65
+
66
+ ## Integration
67
+
68
+ ```python
69
+ from recfold.renderer import render_session
70
+
71
+ render_session(your_decorated_function.last_session)
72
+ ```
73
+
74
+ Alternatively, pass the session returned by `run_traced`, or provide a matching
75
+ `LayoutResult` as the second argument. See [the API guide](renderer.md).
76
+
77
+ The renderer uses only public interfaces from models, analyzer, layout, and
78
+ snapshot. It does not assume an automatic renderer hook in the decorator or
79
+ modify `__init__.py`; imports therefore come from `recfold.renderer` directly.
80
+
81
+ ## Remaining upstream limitations
82
+
83
+ These belong to the existing engine and are documented rather than changed:
84
+
85
+ - The tracer's truncated representation keys can produce false cycle matches.
86
+ - Async/generator execution and arbitrary local-state changes are not traced.
87
+ - Plain recursive methods passed to `run_traced` may bypass the wrapper.
88
+ - Folding relies on representations and observed subtree structure.
89
+ - Large traces are not virtualized, and layout uses straight edges with
90
+ heuristic crossing reduction.
91
+
92
+ The project brief did not include a separate technical specification, so this
93
+ implementation follows the repository's current models and interfaces and the
94
+ team message's zero-dependency Jupyter HTML/SVG requirement.