Jupytext ChangeLog

1.16.2 (2024-05-05)



  • By default, the JupyterLab extension for Jupytext is not included in the build (set HATCH_BUILD_HOOKS_ENABLE=true to include it). This simplifies the installation of Jupytext in pre-commit hooks (#1210)

  • Temporary text notebooks for the --pipe or --check commands are now created in the notebook directory (#1206)

  • Jupytext uses the standard library tomllib in Python 3.11, or tomli in Python 3.10 or older, to match JupyterLab’s dependencies (#1195)

  • The dependencies of the JupyterLab extension were updated (#1216, #1218, #1231)

  • jupytext --sync will not update the timestamp of text notebooks if their content is unchanged (#1215)


  • Jupytext is now tested with pandoc>=3.0. Please note that switching to pandoc>=3.0 will add cell ids to your pandoc:md notebooks (#1006)

1.16.1 (2024-01-13)



  • Fixed an issue about unpairing notebooks from the Jupytext Menu (#1197)

  • JupyterLab’s dependency follow-redirects was updated from 1.15.3 to 1.15.4 (#1203)

1.16.0 (2023-12-03)


  • The Jupytext Menu is back! And text notebooks can be created directly from the launcher. This is an outstanding contribution by Mahendra Paipuri (#1154, #1163). This requires JupyterLab 4.x or Jupyter Notebook 7.x.


  • Jupytext is now configured with pyproject.toml and built with hatch. The layout has been reorganised to follow src-layout (#1140). This is another outstanding contribution by Mahendra Paipuri.

  • The tests are now part of the sdist. They have been reorganized into unit/functional/integration/external (#1167, #1173).

  • The legacy extension for Jupyter Notebook <=6 (the Jupytext Menu) has been removed.


  • The bibliography section in Rmd files does not become a code cell anymore (#1161)

  • Commented code in active-py cells is not uncommented anymore (#1131)

  • The test coverage has been restored (#1167, #1173)

  • We test Jupytext against the pre-release version of JupyterLab, and other dependencies (#1168)

1.15.2 (2023-09-16)


1.15.1 (2023-08-26)


  • We have added a new command line interface jupytext-config that you can use to set Jupytext as the default viewer for text notebooks in Jupyter Lab and Jupyter Notebook 7. Thanks to Thierry Parmentelat for this contribution! (#1094)

1.15.0 (2023-07-30)


  • This version comes with a version of the JupyterLab extension that is compatible with JupyterLab 4.x. Many thanks to Thierry Parmentelat for his PRs! (#1092, #1109)

  • Pandoc 3.0 is now supported, thanks to Raniere Silva for his PR (#1099)

  • We have reorganized the documentation and the README.md (#1031, #1073)

  • Invalid pyproject.toml files will be ignored by Jupytext in Jupyter (#1103)

  • We have updated the pre-commit tools

1.14.7 (2023-06-29)


  • We have updated the GitHub workflows - thanks to Matthew Feickert and to Cristian Le for their help on this subject (#1037)

  • We have removed the upper bound on markdown-it-py<3. Now we test Jupytext with markdown-it-py in versions 2.x and 3.x on the CI (#1075)


  • Notebooks with an empty YAML header work (#1070)

  • Double quote strings in R Markdown options can contain single quotes (#1079)

  • The necessary directories are created when paired notebooks are moved to a sub-folder (#1059)

  • Commented quotes are recognized as such (#1060)

  • Jupytext can use either pkg_resources or packaging to parse version numbers (#1085)

1.14.6 (2023-06-04)


  • This version comes with a build requirement jupyterlab>=3,<4, as the Jupyterlab extension for Jupytext is not compatible with JupyterLab 4 yet (#1054)

  • The JupyterLab extension was released to npm in version 1.3.9.

1.14.5 (2023-02-25)


  • Added Stata as a supported language (#1027) - thanks to Raffaele Mancuso for this contribution

  • Added SAS as a supported language (#1047) - thanks to lawrencet149 for this contribution

  • We have added a series of test to make sure that the main formats support cell tags (#1024)


  • When a metadata key is not a valid identifier, a warning is emitted and the metadata is not saved to the text representation (#1042)

  • The CI was fixed by Matthew Feickert (#1035)

  • We now use concurrency to cancel previous runs on the same branch/PR (#1037)

  • We use both codecov.notify.after_n_builds and comment.after_n_builds to get only the final codecov comment


  • Empty tags are not exported to the text notebook anymore (#960)

  • We updated the yarn.lock file for the jupyter lab extension to address security vulnerabilities (#1030, #1036)

  • In the pre-commit tests we now use main for the main branch

1.14.4 (2022-12-11)


  • Added Wolfram Language as a supported language (#1014) - thanks to Etienne Dechamps for this contribution

1.14.3 (2022-12-11)


  • When the default contents manager is async (i.e. jupyter_server>=2.0.0), the Jupyter server extension for Jupytext derives a contents manager from LargeFileManager instead, as async contents managers are not supported by Jupytext at the moment (#1020)

  • We have made adjustments on the CI as flake8 was moved to GitHub, and Python 3.6 is not available anymore on ubuntu-latest

1.14.2 (2022-11-12)


  • The sample notebooks have been normalized with nbformat.validator.normalize (#1002).

  • The warnings in the test suite that we cannot fix are filtered using a new pytest.ini file

  • We updated the yarn.lock file for the jupyter lab extension to address security vulnerabilities (#984, #1005, #1011)


  • The CI uses Python 3.9 rather than 3.7 when testing conda environments


  • Gnuplot is now supported (#998) - thanks to razimantv for this contribution

  • We now test Jupytext against Python 3.6 to 3.11 on the CI

  • We have added a test to document how to use the folder and prefix matching when pairing notebooks (#974)

1.14.1 (2022-07-29)


  • The timestamp of a paired notebook is the timestamp of the most recent paired file. This fixes the warning “File Changed” after reloading the notebook in Jupyter (#978).

1.14.0 (2022-07-03)


  • The Jupytext configuration file has a new option cm_config_log_level that defaults to info_if_changed. With that value, the contents manager will log a line regarding the configuration file used only when the config file is not the same as the one previously used (#959) - many thanks to R.C. Thomas for suggesting this and thoughtfully testing the patch.

  • Hidden configuration files like .jupytext.toml or .jupytext.py are now ignored by Jupytext’s contents manager when allow_hidden=False (that option was introduced in jupyter_server==2.0.0a1) (#964).

  • We have changed jupytext --set-formats to make it more similar to jupytext --sync. Now --set-formats will not override existing paired files anymore (#969).


  • We have added a test test_pre_commit_hook_sync_with_no_config that documents how to use the pre-commit hook without a configuration file (#967)

1.13.8 (2022-04-04)


  • Text-only notebooks are always trusted (as they don’t include any output cells) (#941)

  • We made sure that our tests also work in absence of a Python kernel (#906)

  • The coverage of the tests folder has been restored at 100%

  • Bash commands like !{cmd} are now correctly escaped in the py:percent format (#938)


  • Added Tcl as a supported language (#930) - thanks to shishitao for this contribution

  • Added Maxima as a supported language (#927) - thanks to Alberto Lusiani for contributing a sample Maxima notebook.


  • The Jupytext contents manager is derived from the LargeFileManager imported from jupyter_server rathen than notebook (#933)

  • Allow for markdown-it-py v2 (#924)

  • We have updated the hooks used in the test pre-commits, to fix an issue on the CI (#940, #942)

  • We updated the yarn.lock file for the jupyter lab extension to address security vulnerabilities (#904, #925, #935, #939)

1.13.7 (2022-02-09)


  • The Jupytext CLI only suggest --update when the target is an .ipynb file (#905) - thanks to st– for this contribution

  • We made sure that commands like cat notebook.md | jupytext --execute work (#908)


  • Added Haskell as supported language (#909) - thanks to codeweber for this contribution


  • We have updated the pre-commit hooks and in particular we switched to the first stable version of black==22.1.0.

  • We require pandoc==2.16.2 for testing. The representation for code cells changed from ``` {.python} to ``` python in that version of Pandoc (#906). We don’t use pandoc>=2.17 in tests at the moment because of the introduction of cell ids that cannot be filtered.

  • Jupytext will not add anymore a UTF-8 encoding on Python scripts when the notebook contains non-ascii characters (#907)

  • We have added pyupgrade to the pre-commit hooks used for developing Jupytext (#907)

1.13.6 (2022-01-11)


  • The text_representation metadata of text notebooks is filtered from .ipynb files both in jupytext.write and in the contents manager for Jupyter (#900)


  • Jupytext will not issue a warning when a format suffix starting with ‘.’, ‘-’ or ‘_’ is passed to the --to option (#901)

1.13.5 (2021-12-27)


  • Jupytext will not open a text notebook that is not UTF-8 (#896)

1.13.4 (2021-12-12)


  • The test suite filters the warnings that don’t belong to Jupytext (#823)


  • The parsing of notebooks that don’t have a YAML header (like docs/formats.md) was improved.

  • The test suite works with pytest-randomly (#838)

1.13.3 (2021-12-04)


  • The “Jupytext Notebook” factory that lets the user configure the Notebook viewer as the default for text notebooks accepts more filetypes: “myst”, “r-markdown” and “quarto” (#803)

  • Empty MyST Markdown files are valid notebooks (#883)

  • Jupytext also works with markdown-it-py v2.0 (#885)

1.13.2 (2021-11-30)


  • The extension for Jupyter Lab benefited from a series of improvements contributed by Frédéric Collonval:

    • A new “Jupytext Notebook” factory offers the option to open text notebooks directly with the notebook view (#803). To use it, follow the instructions in the documentation.

    • The ICommandPalette is optional, for compatibility with RISE within JupyterLab [RISE#605](https://github.com/damianavila/RISE/pull/605)

    • Added support for translation

  • Branch master was renamed to main (links in the documentation were updated)

1.13.1 (2021-10-07)


  • The magic commands in py:percent scripts with no explicit format information remain commented over a round trip (#848)

1.13.0 (2021-09-25)


  • The Jupytext CLI has a new --diff command to show the differences between two notebooks (and if you want to see the changes in a file being updated by Jupytext, use --show-changes) (#799)

  • Jupyter will show the diff between text and ipynb paired notebooks when it cannot open a paired notebook because the ipynb version is more recent. Also, if the inputs in the two files are identical then the notebook will open with no error (#799)

  • The py:percent format will use raw strings when encoding Markdown cells as string, if they contain backslash characters (#836)


  • We have upgraded the jupyterlab extension dependencies and especially ansi-regex to fix a security vulnerability (#857)


  • The Jupytext configuration file is reloaded only when a notebook is opened, saved, or when a different folder is explored (#797)

1.12.0 (2021-09-09)


  • Jupytext supports Quarto notebooks (with .qmd extension) (#837)

  • Jupytext can be configured through the pyproject.toml file. Thanks to Robin Brown for this contribution! (#828)

  • Jupytext now supports OCaml files with .ml extension. Thanks to Quentin Fortier for getting this started (#832)


  • Added more test to make sure that notebooks can be trusted. In practice, notebooks could not be trusted in JupyterLab<3.0.13 because of the absence of cell ids (#826)

1.11.5 (2021-08-31)


  • Fixed typos revealed by codespell - thanks to @hectormz for this contribution (#829)

  • We updated the dependencies of the jupyterlab-jupytext extension to address several security issues (#842) (#843)

  • The Jupytext dev environment (requirements-dev.txt) now uses jupyterlab==3.0.17 rather than 3.0.0 because of another security issue (#839)

1.11.4 (2021-07-14)


  • The documentation illustrates how the cell_markers option (and the other ones) can be set directly in the jupytext.toml config file (#809).

  • The dependency on mdit-py-plugins through markdown-it-py[plugins] was made explicit (#814)


  • System assigns of the form var = !cmd are commented out (#816)

  • Fixed an InconsistentPath issue with notebooks paired with scripts in a folder. The prefix in the Jupytext formats always use /, while paths might use either / or \ (#806)

  • Tests that cannot succeed are skipped when either the Jupytext folder is not a git repository, when sphinx-gallery is too recent, or when pandoc is not up-to-date (#814)

  • Removed the mention of ‘–update’ in ‘jupytext –pipe’ since outputs are preserved already

1.11.3 (2021-06-10)


  • Jupytext CLI has a new option --use-source-timestamp that sets the last modification time of the output file equal to that of the source file (this avoids having to change the timestamp of the source file) (#784)

  • In the pre-commit mode, Jupytext now uses the commit timestamp to determine which file in the pair is the most recent (#780)


  • Dependencies of the JupyterLab extension have been upgraded to fix a security vulnerability (#798)

  • The --warn-only option also applies to pipes. Use this if the pipe may fail, e.g. if you apply black on a possibly invalid script (#781)

  • Variables assigned from a magic command are commented out in py scripts (#781)

  • Fixed a round-trip issue on notebooks that have None/null in their metadata (#792)

1.11.2 (2021-05-02)


  • Jupytext’s dependency markdown-it-py is now in v1 (#769)

  • The optional argument fmt in jupytext.reads now has the default value None - thanks to Yuvi Panda (#763)


  • All text files are opened with an explicit utf-8 encoding (#770)

  • Previously --pipe black was not always putting two blank lines between functions. To fix that we load the internal Jupytext cell metadata like lines_to_next_cell from the text file rather than ipynb (#761)

  • The timestamp of the source file is not updated any more when the destination file is not in the pair (#765, #767)


  • A new test documents when the ipython3 pygment lexer appears in MyST Markdown files (#759)

1.11.1 (2021-03-26)


  • Format options stored in the notebook itself are now taken into account (Fixes #757)

1.11.0 (2021-03-18)


  • The jupytext.toml config file can now be used together with the jupytext pre-commit hook (#752)

  • The notebook_extensions option of the jupytext.toml file now works (#746)


  • The options in jupytext.toml where renamed to match the jupytext metadata in the text notebooks. One should now use formats rather than default_jupytext_formats and notebook_metadata_filter rather than default_notebook_metadata_filter (#753)

1.10.3 (2021-03-07)


  • We have updated marked, an indirect dependency of the jupyterlab-jupytext extension, to fix a moderate vulnerability (#750).

  • We use non-random cell ids in the tests to avoid test failures due to duplicate cell ids (#747)

1.10.2 (2021-02-17)


  • We have adjusted the MANIFEST.in file to exclude the node_modules but still include the JupyterLab extension that was missing in the .tar.gz (and conda) package in v1.10.1. Many thanks to Martin Renou for providing the fix at (#741)

1.10.1 (2021-02-11)


  • The recursive glob pattern **/*.ipynb is now supported by Jupytext - Thanks to Banst for this contribution (#731)

  • Sage notebooks are supported. They can be converted to .sage and .md files and back. Thanks to Lars Franke for suggesting this! (#727)

  • Jupytext is also accessible with python -m jupytext. Thanks to Matthew Brett for his PR! (#739)


  • We have tested Jupytext with the new cell ids introduced in nbformat>=5.1.0. Cell ids are preserved by the --sync and --update command. So we removed the constraint on the version of nbformat (#735).


  • We filtered out the node_modules folder from the .tar.gz package for Jupytext (#730)

1.10.0 (2021-02-04)


  • Jupytext has a pre-commit hook! Many thanks to John Paton and Aaron Gokaslan for making this happen (#698)

  • Jupytext CLI will not rewrite files that don’t change (#698).

  • If you want to see the diff for changed files, use the new --diff option (#722)

  • We have added isort and autoflake8 to the pre-commit configuration file used for developing the Jupytext project (#709)

  • We made sure that py:percent scripts end with exactly one blank line (#682)

  • We checked that Jupytext works well with symbolic links to folders (not files!) (#696)


  • Jupytext does not work properly with the new cell ids of the version 4.5 of nbformat>=5.1.0 yet, so we added the requirement nbformat<=5.0.8 (#715)

  • Jupytext will issue an informative error or warning on notebooks in a version of nbformat that is not known to be supported (#681, #715)


  • Code cells that contain triple backticks (or more) are now encapsulated with four backticks (or more) in the Markdown and MyST Markdown formats. The version number for the Markdown format was increased to 1.3, and the version number for the MyST Markdown format was increased to 0.13 (#712)

  • Indented magic commands are supported (#694)

1.9.1 (2021-01-06)


  • Include the lab extension that was missing in the conda package (#703).

1.9.0 (2021-01-05)


  • The Jupytext extension for JupyterLab is compatible with Jupyter Lab 3.0, thanks to Martin Renou’s awesome contribution (#683).

1.8.2 (2021-01-04)


  • Jupytext 1.8.2 depends on python>=3.6. The last version of Jupytext explicitly tested with Python 2.7 and 3.5 was Jupytext 1.7.1 (#697).

1.8.1 (2021-01-03)


  • The dependency on markdown-it-py is conditional on python>=3.6 (#697)

1.8.0 (2020-12-22)


  • Removed support for Python 2.7 and 3.5, a preliminary step towards a JupyterLab 3.0-compatible extension (#683)

  • The MyST Markdown format uses markdown-it-py~=0.6.0 (#692)

1.7.1 (2020-11-16)


  • Text notebooks have the same format and mimetype as ipynb notebooks. This fixes the File Load Error - content.indexOf is not a function error on text notebooks (#659)

1.7.0 (2020-11-14)


  • Jupytext’s contents manager uses the parent CM’s get and save methods to read and save text files, and explicitly calls jupytext.reads and jupytext.writes to do the conversion. We don’t use mock nor internal parent methods any more. Thanks to Max Klein for helping making this work! (#634, #635)

  • Thanks to the above, Jupytext can work on top of contents manager that don’t derive from FileContentsManager, and in particular it works with jupyterfs (#618)

  • The documentation was reorganized. README.md was simplified and now includes many links to the documentation.

  • The documentation now uses myst_parser rather than recommonmark. And we use conda on RTD (#650, #652)

  • The readf and writef functions were dropped (they had been deprecated in favor of read and write in June 2019, v1.2.0)

  • The description & dependencies of the JupyterLab extension were updated (#654)

  • The --set-kernel - command, on a Python notebook, gives an explicit error when no kernel is not found that matches the current Python executable.

  • All the GitHub workflow files were concatenated into a unique file, and we have added an pypi-publish step to automatically publish the package on PyPi when new releases are created.

  • The CHANGELOG.md file was moved under docs to better expose the history of changes.


  • Configuration errors are reported in the console and/or in Jupyter (#613)

  • Jupytext’s Contents Manager internal errors are logged on the console, and trigger an HTTP Error 500 (#638)

  • The GitHub actions run on both push events and pull requests, and duplicate jobs are skipped (#605)

  • Jupytext has a tox.ini file, thanks to Chris Sewell (#605)

  • Jupytext is tested against Python 3.9

  • The execution cell metadata is now filtered by default (#656)


  • Optional dependency on sphinx-gallery frozen to version ~=0.7.0 (#614)

  • Codecov/patch reports should be OK now (#639)

  • Jupytext tests work on non-English locales (#636)

  • Cell metadata that are already present in text notebook can be filtered out using a config file (#656)

  • Optional cell attributes like attachments are preserved (#671)

1.6.0 (2020-09-01)


  • New option hide_notebook_metadata to encapsulate the notebook metadata in an HTML comment (#527)

  • New option root_level_metadata_as_raw_cell. Set it to False if you don’t want to see root level metadata of R Markdown notebooks as a raw cell in Jupyter (#415)

  • New option doxygen_equation_markers to translate Markdown equations into Doxygen equations (#517)

  • New option custom_cell_magics to comment out cells starting with user-specific cell magics (#513)

  • Documented how to use isort on notebooks (#553)

  • jupytext notebook.ipynb --to filename.py will warn that --to is used in place of --output.

  • jupytext --set-formats filename.py will suggest to use --sync instead of --set-formats (#544)

  • Warn if ‘Include Metadata’ is off when saving text files in Jupyter (#561)

  • Test that notebooks paired through a configuration file are left unmodified (#598)

  • Test that metadata filters in the configuration files are taken into account when using jupytext --to (#543)

  • New argument --run-path to execute the notebooks at the desired location (#595)

  • Activated GitHub code scanning alerts


  • Jupytext now depends on markdown-it-py (Python 3.6 and above) and always features the MyST-Markdown format, thanks to Chris Sewell (#591)

  • The md:myst and md:pandoc are always included in the Jupytext formats, and an informative runtime error will occur if the required dependencies, resp. markdown-it-py and pandoc, are not installed. (#556)

  • The # %% cell marker has the same indentation as the first line in the cell (#562)

  • Jupytext is now installed from source on MyBinder to avoid cache issues (#567)

  • The tests that execute a notebook are now skipped on Windows to avoid timeout issues (#489)


  • Only scripts can have an encoding comment, not Markdown or R Markdown files (#576)

  • Spaces in --pipe commands are supported (#562)

  • Bash commands starting with special characters are now correctly detected, thanks to Aaron Gokaslan (#587)

  • MyST Markdown files are recognized as such even if MyST-Markdown is not available (#556)

  • Build JupyterLab with dev-build=False and minimize=False on mybinder to avoid build errors

  • Configured coverage targets in codecov.yml

1.5.2 (2020-07-21)


  • The documentation uses the Alabaster theme


  • Preserve indentation when commenting out magic commands (#437)

  • Fixed MyBinder - jupytext.py is not a configuration file (#559, #567)

1.5.1 (2020-07-05)


  • Added toml as a dependency (#552).

  • Filtered out __pycache__ and .pyc files from the pip package.

  • Fixed coverage upload by adding coverage as a dependency to the conda CI workflow.

  • Fixed the conda CI / Python 2.7 job by explicitly installing backports.functools_lru_cache (#554).

1.5.0 (2020-06-07)


  • Jupytext can use a local or global configuration file (#508)

  • Jupytext can pair notebooks in trees. Use e.g. notebooks///ipynb,scripts///py:percent if you want to replicate the tree of notebooks under notebooks in a folder named scripts (#424)

  • The extension for Jupyter Notebook has a New Text Notebook menu that creates text-only notebooks (#443)

  • Groovy and Java are now supported, thanks to Przemek Wesołek (#500)

  • The Coconut language is also supported, thanks to Thurston Sexton (#532)

  • Resource files with .resource extension from the Robot Framework are supported, thanks to Hiski Valli (#535)

  • Jupytext is tested in pip and conda environments, on Linux, Mac OS and Windows, using Github actions (#487)

  • Jupytext uses pre-commit checks and automatic reformatting with pre-commit, black and flake8 (#483)

  • Documentation improvements:

    • Mention that the YAML header can be created with either --set-kernel, --set-formats, or both (#485)

    • Mention that one should use double quotes, not single quotes, around jupytext --check commands like "pytest {}" on Windows (#475)

    • Improved error message when a file is in a version that can’t be read by Jupytext (#531)


  • Skip the jupytext --execute tests when the warning Timeout waiting for IOPub output occurs, which is the case intermittently on Windows (#489)

  • Fixed wrong paired paths when syncing with the –pre-commit flag (#506)

1.4.2 (2020-04-05)


  • Added an example with custom notebook metadata (#469)

  • Added an example with custom cell tags (#478)


  • The outputs from the .ipynb file are matched with the input cells from the text file with less strict rules. In this version, a search and replace on the text file will not remove the outputs any more (#464).

  • Update parsing of myst notebooks to new (markdown-it based) parser (please upgrade to myst-parser to version ~0.8) (#473)

  • Use os.path.samefile when searching for the kernel that corresponds to the current environment (--set-kernel -)


  • Fixed the CLI example for not commenting out magic commands: --opt comment_magics=false. In addition, most of the jupytext commands in using-cli.md are now tested! (#465)

  • jupytext.read and jupytext.write now give more meaningful errors when the format information is incorrect (#462)

  • Multiline comments starting or ending with quadruple quotes should not cause issues anymore (#460)

  • Fixed active cells in the py:percent format (#477)

1.4.1 (2020-03-19)


  • Script of script (SoS) notebooks are now supported. Thanks to Thomas Pernet-coudrier for contributing the sample notebook (#453).

  • New MyST Markdown format (md:myst), developed by the ExecutableBookProject team. Read more about the MyST Markdown format in the documentation. And many thanks to Chris Sewell for contributing the actual implementation! (#447 #456 #458)


  • When using jupytext --pipe cmd, the output of cmd should not appear in the terminal (#432)

1.4.0 (2020-03-09)


  • The new jupyterlab extension (in version 1.2.0) is compatible with JupyterLab 2.0. Many thanks to Jean Helie! (#449)

  • It is not compatible with JupyterLab 1.x anymore. If you wish, you can install manually the previous version of the extension with jupyter labextension install jupyterlab-jupytext@1.1.1.


  • Display the output/errors of command executed with jupytext --pipe or jupytext --check (#432)

1.3.5 (2020-03-08)


  • Removed leading slash in notebook paths (#444)

  • Fixed jupytext --set-formats when using formats with prefix and/or suffix (#450)

1.3.4 (2020-02-18)


  • C# and F# Jupyter notebooks are now supported (#427, #429)


  • jupytext --to script *.ipynb now computes the script extension for each notebook (#428)

  • Fix shebang handling for languages with non-# comments, by Jonas Bushart (#434)

  • Indented bash commands are now commented out (#437)

  • The main formats are documented in jupytext --help (#426, #433)

1.3.3 (2020-01-27)


  • Jupytext has a logo! Many thanks to Kyle Kelley for contributing the actual logo (#423), and to Chris Holdgraf for suggesting this (#260).

  • Nested metadata filtering is now supported! You can use this to rid of jupytext_version if you wish (#416).


  • Code cells in the Markdown format can contain triple backticks inside multiline strings (#419).

  • Changes in the YAML header when running jupytext --test on text files are ignored (#414).

1.3.2 (2020-01-08)


  • The --pre-commit mode now ignores non-notebook files in the index (#338).

  • nbformat_minor is taken from the notebook with outputs When inputs and outputs are merged.

1.3.1 (2019-12-26)


  • New nomarker format in the Jupyter Notebook and JupyterLab extensions (#397)


  • The normarker format replaces the one previously named bare.


  • Multiline strings are now accepted in the YAML header (#404). Fix contributed by ZHUO Qiang (#405).

  • Fixed the instructions on how to use multiline comments for all Markdown cells in the py:percent format, by ZHUO Qiang (#403).

  • Added cd to the list of magic commands.

  • Do not read paired files when --set-formats is being used (fixes #399).

1.3.0 (2019-11-23)

See also What’s new in Jupytext 1.3?


  • Pairing a notebook to both .md and .py is now supported. Input cells are loaded from the most recent text representation (#290)

  • Both the Jupyter Notebook and the Jupyter Lab extension for Jupytext were updated (#290)

  • Raw cells are now encoded using HTML comments (<!-- #raw --> and <!-- #endraw -->) in Markdown files (#321)

  • Markdown cells can be delimited with any of <!-- #region -->, <!-- #markdown --> or <!-- #md --> (#344)

  • Code blocks from Markdown files, when they don’t have an explicit language, appear in Markdown cells in Jupyter (#321)

  • Code blocks with an explicit language and a .noeval attribute are inactive in Jupyter (#347)

  • Markdown and raw cells can be quoted with triple quotes in the py:percent format. And Markdown cells can start with just # %% [md] (#305)

  • The light format uses # + [markdown] rather than the previous cell_type metadata to identify markdown cells with metadata (#356)

  • Explicit Markdown cells in the light format # + [markdown] can use triple quotes (#356)

  • IPython magic help commands like float? are classified as magics, and thus commented in Python scripts (#297)

  • Cell metadata can be encoded as either key=value (the new default) or in JSON. An automatic option cell_metadata_json should help minimize the impact on existing files (#344)

  • R Markdown hidden inputs, outputs, or cells are now mapped to the corresponding Jupyter Book tags by default (#337)

  • The notebook metadata filter is automatically updated when extra keys are added to the YAML header (#376)

  • The Jupyter Notebook extension for Jupytext is compatible with Jupyter Notebook 6.0 (#346)

  • jupytext notebook.py --to ipynb updates the timestamp of notebook.py so that the paired notebook still works in Jupyter (#335, #254)

  • jupytext --check pytest notebook.ipynb can be used to run test functions in a notebook (#286)

  • jupytext --check and jupytext --pipe can run commands that only operate on files: when {} is found in the text of the command, jupytext saves the text representation of the notebook in a temp file, and replaces {} with the name of that file before executing the command. (#286)

  • Documented how to sync notebooks in a pre-commit hook (#338)

  • Added support for Rust/Evxcr, by Jonas Bushart (#351)

  • Added support for Robot Framework, by Asko Soukka (#378)

  • Added support for PowerShell (#349)


  • Use CHANGELOG.md rather than HISTORY.rst


  • Commands like cat = x are not magic commands, so they are not commented any more (#339)

  • Fixed an inconsistent round trip (code cell with "cat" being converted to a markdown cell) in the py:light format (#339)

  • jupytext --test textfile.ext now really compares the text file to its round trip (rather than the corresponding notebook) (#339)

  • Markdown cells that contain code are now preserved in a round trip through the Markdown and R Markdown formats (#361)

  • Code cells with a %%python3 cell magic are now preserved in a round trip through the Markdown format (#365)

  • jupytext --execute runs the notebook in its folder (#382)

  • Strings in the metadata of code cells are quoted in the Rmd representation. And we escape R code in chunk options with #R_CODE# (#383)

1.2.4 (2019-09-19)


  • The documentation includes a mention on how to set metadata filters at the command line (#330)

  • Jupytext will not catch any error when the flag --warn-only is not set (#327)


  • Now the flag --warn-only catches every possible error (#263)

  • .md and .markdown files are treated identically (#325)

  • Fixed --set-kernel when using pipes (#326)

  • Fixed utf-8 encoding on stdout on Python 2 (#331)

1.2.3 (2019-09-02)


  • Dependency on setuptools in pandoc.py made optional to fix the build of the conda package (#310, #323)

1.2.2 (2019-09-01)


  • Documentation includes a section on how to use Jupytext as a Python library (#317)

  • Mention of the server extension in the documentation (#304)

  • Text notebooks can be tested with jupytext --execute notebook.md (#303)

  • The default value of as_version in jupytext.read is nbformat.NO_CONVERT, as for nbformat.read

  • Jupytext tests are now included in sdist (#310)


  • Fixed the usability of the fmt argument in jupytext.read (#312)

  • Fixed the download notebook error when c.notebook_extensions has a custom value (#318)

  • String delimiters in commented text are now ignored (#307)

1.2.1 (2019-07-20)


  • Added documentation on how to use Jupytext with JupyterLab 0.35 (#299)

  • and on using Jupytext with the pre-commit package manager (#292)

  • The read and write functions are easier to use (#292)


  • Jupytext now ships the jupyterlab-jupytext extension in version 1.0.2. The version 1.0.1 erroneously introduces a target_formats metadata in the jupytext section, instead of formats, and works only after two clicks.

1.2.0 (2019-07-18)


  • New --execute option in Jupytext CLI (#231)

  • The --set-formats option in Jupytext CLI also triggers --sync, allowing shorter commands.

  • jupytext’s read and write functions can be used as drop-in replacements for nbformat’s ones (#262).

  • jupytext --sync will now skip unpaired notebooks (#281).

  • The JupyterLab extension was updated. It now works on on text files (#213) and has a new option to include (or not) the metadata in the text representation of the notebook.

  • Jupytext’s contents manager class is derived dynamically from the default CM class for compatibility with jupyter_server (#270)

  • Removed dependency on mock and testfixtures, thanks to Jean-Sebastien Dieu (#279)

  • Added support for .markdown extension (#288)


  • The jupyterlab-jupytext extension shipped with the python package is in version 1.0.1, and is compatible only with JupyterLab >= 1.0. If you use an earlier version of JupyterLab, please install the version 0.19 of the extension with jupyter labextension install jupyterlab-jupytext@0.19 (#276, #278)

  • Text files can be unpaired (#289)

1.1.7 (2019-06-23)


  • Added support for Scala notebook, by Tobias Frischholz (#253)

  • Added a getting started notebook for jupytext (and Binder), by Chris Holdgraf (#257)

  • The Markdown and R Markdown representations are now tested for all the languages

  • The Jupytext notebook extension also works when the notebook is a text file (#213)


  • The Jupytext Menu in Jupyter Notebook is now compatible with jupyter_nbextensions_configurator (#178)

  • Entries in the Jupytext menu are updated when the menu is hovered on (#248)

  • Fixed link to .md files in the documentation (#255)

1.1.6 (2019-06-11)


  • Jupytext now supports Javascript and Typescript, thanks to Hatem Hosny (#250)

  • Jupytext works with Python 3.8 as well


  • Fix global auto pairing (#249)

1.1.5 (2019-06-06)


  • Fixed implicit dependency on jupyter_client (#245)

1.1.4 (2019-06-05)


  • New argument --set-kernel in Jupytext command line (#230)

  • Jupytext now accepts --to script or --to auto (#240)

  • Jupytext now has a real Sphinx documentation on readthedocs, thanks to Chris Holdgraf (#237)


  • Invalid notebooks may cause a warning, but not a fatal error (#234)

  • Jupyter server extension leaves the contents manager unchanged if it is a sub-class of Jupytext’s CM (#236)

  • Fixed format inference when metadata is present but not format information (#239)

  • Preserve executable and encoding information in scripts with metadata (#241)

1.1.3 (2019-05-22)


  • Support for IDL notebooks and .pro scripts (#232)

1.1.2 (2019-05-16)


  • Jupytext’s content manager has a new notebook_extensions option (#224, #183)

  • Cells can be made inactive in scripts with the active-ipynb cell tag (#226)


  • Directories ending in .jl (or .ipynb) are not notebooks (#228)

  • Empty notebooks have no language (#227)

1.1.1 (2019-04-16)


  • Jupytext server extension leaves the contents manager unchanged when it is already a subclass of TextFileContentsManager (#218)

  • The base class for TextFileContentsManager defaults to FileContentsManager when LargeFileManager is not available (#217)

1.1.0 (2019-04-14)


  • Markdown and R Markdown formats now support metadata (#66, #111, #188)

  • The light format for Scripts can use custom cell markers, e.g. Vim or VScode/PyCharm folding markers (#199)

  • Pandoc’s Markdown format for Jupyter notebooks is available in Jupytext as md:pandoc (#208)


  • Jupytext’s contents manager is now based on LargeFileManager to allow large file uploads (#210)

  • YAML header parsed with yaml.safe_load rather than yaml.load (#215)

  • IPython line magic can be split across lines (#209)

  • jupytext --to py rather than --to python in the README (#216)

1.0.5 (2019-03-26)


  • Fix the error ‘notebook file has changed on disk’ when saving large notebooks (#207)

1.0.4 (2019-03-20)


  • Wildcard are now supported on Windows (#202)

  • Trusted notebooks remain trusted when inputs cells are modified (#203)


  • Pre-commit mode adds the result of conversion to the commit (#200)

1.0.3 (2019-03-13)


  • Matlab and Octave notebooks and scripts are now supported (#197)


  • notebook_metadata_filter = "all" now works (#196)

  • Default pairing in subfolders fixed in Jupyter Lab (#180)

1.0.2 (2019-02-27)


  • Rename notebooks in pairs in the tree view (#190)

  • Associate .scm file extension with Scheme scripts (#192)

  • Added support for Clojure, by bzinberg (#193)


  • Allow spaces between ? or ! and python or bash command (#189)

1.0.1 (2019-02-23)


  • Exclude tests in package deployment (#184)

  • Jupytext’s serverextension only runs selected init steps (#185)

  • Added an additional test for magic arguments (#111)

1.0.0 (2019-02-19)


  • Jupytext now includes a Jupyter Notebook extension and a JupyterLab extension (#86).

  • Jupytext command line has more arguments: --paired-paths to list the paths for the paired representations of the notebook, and --sync to synchronise the content of all paired paths based on the most recent file (#146). In addition, the --from argument is optional even when the notebook is read from stdin (#148).

  • The pairing information, and more generally the notebook metadata can be edited with the CLL, see the --set-formats and the --update-metadata arguments (#141).

  • Jupytext can --pipe the text representation of a notebook to external programs like black or flake8 (#154, #142)

  • The Python representation of notebooks containing PEP8 cells is now expected to be PEP8 compliant (#154).

  • Format specification allow prefix and suffix for path and file name (#138, #142). Use ipynb,prefix/suffix.py:percent to pair the current notebook named notebook.ipynb to a script named prefixnotebooksuffix.py. Suffix and prefix can also be configured on the ipynb file, with the same syntax.

  • Introducing a new hydrogen format for scripts, which derives from percent. In that format Jupyter magic commands are not commented (#59, #126, #132).

  • Introducing a new bare format for scripts, which derives from light. That format has no cell marker. Use a notebook metadata filter {"jupytext": {"notebook_metadata_filter":"-all"}} if you want no YAML header (#152).

  • The default format for R script is now light, as for the other languages.

  • Added support for q/kdb+ notebooks (#161).

  • Python scripts or Markdown documents that have no Jupyter metadata receive a metadata filter that ensures that metadata is not exported back to the text representation (#124).

  • Metadata filters are represented as strings rather than dictionaries to make YAML headers shorter. Previous syntax from #105 is still supported. They were also renamed to notebook_metadata_filter and cell_metadata_filter.

  • Markdown and RMarkdown formats have a new option split_at_heading to split Markdown cells at heading (#130)


  • Main language of scripts is inferred from script extension. Fixes a round trip conversion issue for Python notebooks with a Javascript cell.

  • Non-Python scripts opened as notebooks in Jupyter are now correctly saved even when no matching kernel is found.

  • Jupyter magic commands like ls are commented in the light and R markdown format (#149).

  • Cell starting with %%html, %%latex are now commented out in the light, percent and Rmd formats (#179).

0.8.6 (2018-11-29)


  • The language_info section is not part of the default header any more. Language information is now taken from metadata kernelspec.language. (#105).

  • When opening a paired notebook, the active file is now the file that was originally opened (#118). When saving a notebook, timestamps of all the alternative representations are tested to ensure that Jupyter’s autosave does not override manual modifications.

  • Jupyter magic commands are now commented per default in the percent format (#126, #132). Version for the percent format increases from ‘1.1’ to ‘1.2’. Set an option comment_magics to false either per notebook, or globally on Jupytext’s contents manager, or on jupytext’s command line, if you prefer not to comment Jupyter magics.

  • Jupytext command line has a pre-commit mode (#121).

0.8.5 (2018-11-13)


  • bash scripts as notebooks (#127)

  • R scripts with .r extension are supported (#122)

  • Jupytext selects the first kernel that matches the language (#120)

0.8.4 (2018-10-29)


  • Notebook metadata is filtered - only the most common metadata are stored in the text representation (#105)

  • New config option freeze_metadata on the content manager and on the command line interface (defaults to False). Use this option to avoid creating a YAML header or cell metadata if there was none initially. (#110)

  • Language magic arguments are preserved in R Markdown, and also supported in light and percent scripts (#111, #114, #115)

  • First markdown cell exported as a docstring when using the Sphinx format (#107)

0.8.3 (2018-10-19)


  • Frozen cells are supported in R Markdown, light and percent scripts (#101)

  • Inactive cells extended to percent scripts (#108)

  • jupytext gains a --version argument (#103)

  • “ExecuteTime” cell metadata is not included in the text representation anymore (#106)

0.8.2 (2018-10-15)


  • Round trip conversion testing with jupytext --test was improved (#99)

  • Round trip conversion tested on Jake Vanderplas’ Python for Data Science Handbook.


  • Nested lists and dictionaries are now supported in notebook metadata

  • Final empty code cell supported in Sphinx representation

0.8.1 (2018-10-11)


  • Sphinx format tested on World population notebook (#97)

  • Mirror test made stronger on this occasion!

  • Markdown representation recognize Julia, Scheme and C++ code cells as such

  • Light representation of Scheme and C++ notebooks fixed (#61)

0.8.0 (2018-10-10)


  • All jupytext related metadata goes to a jupytext section (#91). Please make sure your collaborators use the same version of Jupytext, as the new version can read previous metadata, but not the opposite.

  • Notebooks extensions can be prefixed with any prefix of at most three chars (#87).

  • Export of the same notebook to multiple formats is now supported. To export to all python formats, plus .ipynb and .md, use "jupytext": {"formats": "ipynb,pct.py:percent,lgt.py:light,spx.py:sphinx,md"},.

  • README includes a short section on how to extend light and percent formats to more languages (#61).

  • Jupytext’s contents manager accepts the auto extension in default_jupytext_formats (#93).

  • All Jupyter magics are escaped in light scripts and R markdown documents. Escape magics in other formats with a comment_magics metadata (true or false), or with the contents manager comment_magics global flag (#94).


  • Trusting notebooks made functional again.

  • Command line jupytext returns a meaningful error when no argument is given.

  • Fixed global pairing configuration (#95).

0.7.2 (2018-10-01)


  • light and percent formats made available for scheme and cpp notebooks. Adding more formats is straightforward - just add a new entry to _SCRIPT_EXTENSIONS in languages.py, a sample notebook and a mirror test (#61)

  • Format name is automatically appended to extension in jupytext_formats when notebook is loaded/saved.


  • Notebooks extensions can only be prefixed with .nb (#87)

0.7.1 (2018-09-24)


  • Markdown cells header in sphinx gallery format may have a space between first # and following.

0.7.0 (2018-09-23)


  • Header for cells in percent format is more robust: use [markdown] and [raw] to identify cell types. Cell type comes after the cell title. (#59)

  • Jupytext can read and write notebooks as Hydrogen/VScode/Spyder/PyCharm compatible scripts (cells starting with # %%) (#59)

  • Jupytext can read and write notebooks as Sphinx-gallery compatible scripts (#80)

  • Metadata are supported for all cell types in light python and percent formats (#66). Due to this, light python format version is now 1.3. Light python notebooks in versions 1.1 and 1.2 are still readable.

  • Command line jupytext has a from argument, and now accepts notebook from the standard input.


  • Fix merging of input and output notebooks (#83)

  • Removed extra new line on stdout in command line jupytext (#84)

0.6.5 (2018-09-13)


  • Code lines that start with a quotation mark in Jupyter are commented in the corresponding Python and Julia scripts (#73)

  • Update pypy, add flake8 tests on Travis CI (#74)


  • Import notebook.transutils before notebook.services.contents.filemanager (#75)

0.6.4 (2018-09-12)


  • Jupytext will not load paired notebook when text representation is out of date (#63)

  • Package tested against Python 3.7 (#68)


  • Allow unicode characters in notebook path (#70)

  • Read README.md as unicode in setup.py (#71)

0.6.3 (2018-09-07)


  • Lighter cell markers for Python and Julia scripts (#57). Corresponding file format version at 1.2. Scripts in previous version 1.1 can still be opened.

  • New screenshots for the README.


  • Command line conversion tool jupytext fixed on Python 2.7 (#46)

0.6.2 (2018-09-05)


  • Initial support for Jupyter notebooks as Julia scripts (#56)

  • Command line conversion tool jupytext has explicit to and output options (#46)

  • Round trip test with jupytext --test improved (#54)

  • Improved README (#51)


  • testfixtures now in requirements (#55)

  • Empty code cells are now preserved (#53)

0.6.1 (2018-08-31)


  • Package and conversion script renamed from nbrmd to jupytext.

0.6.0 (2018-08-31)


  • Cell parsing and exporting done in two specialized classes. This is way easier to read. Pylint score at 9.9 !

  • Python file format updated to 1.1: default end of cell for python scripts is one blank space. Two blank spaces are allowed as well. Now you can reformat code in Python IDE without breaking notebook cells (#38).

  • Added support for plain markdown files (#40, #44).

  • Demonstration notebooks more user friendly (#45).

  • Command line tool simpler to use (#46).

  • Start code patterns present in Jupyter cells are escaped.

  • Default nbrmd_format is empty (mwouts/nbsrc/#5): no Jupyter notebook is created on disk when the user opens a Python or R file and saves it from Jupyter, unless the users asks for it by setting nbrmd_format.


  • Fixed message in the nbsrc script (#43)

  • Technical metadata don’t appear any more in scripts unless required (#42)

  • Code cells that are fully commented remain code cells after round trip (#41)

0.5.4 (2018-08-24)


  • R to Rmd conversion compares well to knitr::spin (#26)

  • Increased coverage to 98%

0.5.3 (2018-08-22)


  • Read and write version to the same metadata (#36)

0.5.2 (2018-08-22)


  • Classical jupyter extensions (autoreload, rmagics) are also escaped (#35)

  • Explicit file format version, set at 1.0, to avoid overriding ipynb files by accident (#36)

0.5.1 (2018-08-21)


  • Source only notebooks can be trusted.

0.5.0 (2018-08-21)


  • Jupyter magic commands escaped when exported (#29)

  • ‘endofcell’ option for explicit (optional) end-of-cell marker (#31)

  • ‘active’ cell option now supported for .py and .R export (#30)

  • Raw cells now preserved when exported to .py or .R (#32)

  • Extensions can be prefixed, like .nb.py, (mwouts/nbsrc#5)

  • When a file with an extension not associated to ‘ipynb’ is opened and saved, no ‘ipynb’ file is created (mwouts/nbsrc#5)

  • Extensions can now be a sequence of groups. For instance, nbrmd_formats="ipynb,nb.py;script.ipynb,py" will create an ipynb file when a nb.py is opened (and conversely), and a script.ipynb file when a py file is opened (mwouts/nbsrc#5)

  • nbsrc script was moved to the nbrmd package. The nbsrc package now only contains the documentation (mwouts/nbsrc#3)

0.4.6 (2018-07-26)

  • Ping pypi - previous version still not available

0.4.5 (2018-07-26)


  • Removed dependency of setup.py on yaml

0.4.4 (2018-07-26)


  • Package republished with python setup.py sdist bdist_wheel to fix missing dependencies

0.4.3 (2018-07-26)


  • Multiline comments now supported #25

  • Readme refactored, notebook demos available on binder #23


  • ContentsManager can be imported even if notebook.transutils is not available, for compatibility with older python distributions.

  • Fixed missing cell metadata #27

  • Documentation tells how to avoid creating .ipynb files #16

0.4.2 (2018-07-23)


  • Added test for R notebooks

  • Added pylint badge, imports now in correct order

  • New active cell metadata that allows cell activation only for desired extensions (currently available for Rmd and ipynb extensions only)

0.4.1 (2018-07-20)


  • Indented python code will not start a new cell #20

  • Fixed parsing of Rmd cell metadata #21

0.4.0 (2018-07-18)


  • .py format for notebooks is lighter and pep8 compliant


  • Default nbrmd config not added to notebooks (#17)

  • nbrmd_formats becomes a configurable traits (#16)

  • Removed nbrmd_sourceonly_format metadata. Source notebook is current notebook when not .ipynb, otherwise the first notebook format in nbrmd_formats (not .ipynb) that is found on disk

0.3.0 (2018-07-17)


  • Introducing support for notebooks as python .py or R scripts .R

0.2.6 (2018-07-13)


  • Introduced nbrmd_sourceonly_format metadata

  • Inputs are loaded from .Rmd file when a matching .ipynb file is opened.


  • Trusted notebooks remain trusted (#12)

0.2.5 (2018-07-11)


  • Outputs of existing .ipynb versions are combined with matching inputs of R markdown version, as suggested by @grst (#12)


  • Support for unicode text in python 2.7 (#11)

0.2.4 (2018-07-05)


  • nbrmd will always open notebooks, even if header of code cells are not terminated. Merge conflicts can thus be solved in Jupyter directly.

  • New metadata ‘main language’ that preserves the notebook language.


  • dependencies included in setup.py

  • pre_save_hook work with non-empty notebook_dir (#9)

0.2.3 (2018-06-28)


  • Screenshots in README


  • RMarkdown exporter for nbconvert fixed on non-recent python

  • Tests compatible with other revisions of nbformat >= 4.0

  • Tests compatible with older pytest versions

0.2.2 (2018-06-28)


  • RMarkdown exporter for nbconvert

  • Parsing of R options robust to parenthesis

  • Jupyter cell tags are preserved


  • requirements.txt now included in pypi packages

0.2.1 (2018-06-24)


  • Support for editing markdown files in Jupyter

  • New pre-save hook update_selected_formats that saves to formats in metadata ‘nbrmd_formats’

  • Rmd cell options directly mapped to cell metadata


  • ContentManager compatible with Python 2.7

0.2.0 (2018-06-21)


  • The package provides a RmdFileContentsManager for direct edit of R markdown files in Jupyter

  • Notebook metadata and cell options are preserved

0.1.1 (2018-06-19)


  • nbrmd prints the result of conversion to stdout, unless flag -i is provided

  • Notebooks with R code chunks are supported

0.1 (2018-06-18)

  • Initial version with the nbrmd converter and Jupyter pre_save_hook