Skip to main content

A Python library for writing pandoc filters for tables with batteries included.

Project description

Date: November 30, 2021
docs Documentation Status image2

GitHub Actions Coverage Status

Codacy Code Quality Status Scrutinizer Status CodeClimate Quality Status


Supported versions Supported implementations PyPI Wheel

PyPI Package latest release GitHub Releases Development Status Downloads

Commits since latest release License

conda-forge Conda Recipe Conda Downloads Conda Version Conda Platforms
cite DOI


Pantable is a Python library that maps the pandoc Table AST to an internal structure losslessly. This enables writing pandoc filters specifically manipulating tables in pandoc.

This also comes with 3 pandoc filters, pantable, pantable2csv, pantable2csvx.

pantable is the main filter, introducing a syntax to include CSV table in markdown source. It supports all table features supported by the pandoc Table AST.

pantable2csv complements pantable, is the inverse of pantable, which convert native pandoc tables into the CSV table format defined by pantable. This is lossy as of pandoc 2.11+, which is supported since pantable 0.13.

pantable2csvx (experimental, may drop in the future) is similar to pantable2csv, but introduces an extra column with the fancy-table syntax defined below such that any general pandoc Table AST can be losslessly encoded in CSV format.

Some example uses are:

  1. You already have tables in CSV format.
  2. You feel that directly editing markdown table is troublesome. You want a spreadsheet interface to edit, but want to convert it to native pandoc table for higher readability. And this process might go back and forth.
  3. You want lower-level control on the table and column widths.
  4. You want to use all table features supported by the pandoc’s internal AST table format, which is not possible in markdown for pandoc (as of writing.)

A word on support

Note that the above is exactly how I use pantable personally. So you can count on the round-trip losslessness. pantable and pantable2csv should have robust support since it has been used for years. But since pandoc 2.11 the table AST has been majorly revised. Pantable 0.13 added support for this new AST by completely rewriting pantable, at the same time addresses some of the shortcoming of the original design. Part of the new design is to enable pantable as a library (see Pantable as a library below) so that its functionality can be extended, similar to how to write a pandoc filter to intercept the AST and modify it, you can intercept the internal structure of PanTable and modify it.

However, since this library is completely rewritten as of v0.13,

  • pantable and pantable2csv as pandoc filters should be stable
    • there may be regression, please open an issue to report this
  • round-trip losslessness may break, please open an issue to report this
  • pantable2csvx as pandoc filter is experimental. API here might change in the future or may be dropped completed (e.g. replaces by something even more general)
  • Pantable as a library also is experimental, meaning that the API might be changed in the future.



To manage pantable using pip, open the command line and run

  • pip install pantable to install
    • pip install to install the in-development version
  • pip install -U pantable to upgrade
  • pip uninstall pantable to remove

You need a matching pandoc version for pantable to work flawlessly. See Supported pandoc versions for details. Or, use the Conda method to install below to have the pandoc version automatically managed for you.


To manage pantable with a matching pandoc version, open the command line and run

  • conda install -c conda-forge pantable to install
  • conda update pantable to upgrade
  • conda remove pantable to remove

You may also replace conda by mamba, which is basically a drop-in replacement of the conda package manager. See mamba-org/mamba: The Fast Cross-Platform Package Manager for details.

Note on versions

Supported Python versions

pantable v0.12 drop Python 2 support. You need to pip install pantable<0.12 if you need to run it on Python 2.

To enforce using Python 3, depending on your system, you may need to specify python3 and pip3 explicitly.

Check the badge above or for supported Python versions, further indicates support of pypy in additional of CPython.

Supported pandoc versions

pandoc versioning semantics is MAJOR.MAJOR.MINOR.PATCH and panflute’s is MAJOR.MINOR.PATCH. Below we shows matching versions of pandoc that panflute supports, in descending order. Only major version is shown as long as the minor versions doesn’t matter.

Version Matching [1]
pantable panflute version supported pandoc versions supported pandoc API versions
0.14.1 2.1.3–2.16.x 1.22–1.22.1
0.14 2.1—2.14.x 1.22
0.13 2.0—2.11.x 1.22
not supported 2.10 1.21
0.12 1.12 2.7-2.9 1.17.5–1.20

Note: pandoc 2.10 is short lived and 2.11 has minor API changes comparing to that, mainly for fixing its shortcomings. Please avoid using pandoc 2.10.

To use pantable with pandoc < 2.10, install pantable 0.12 explicitly by pip install pantable~=0.12.4.

Pantable as pandoc filters


This allows CSV tables, optionally containing markdown syntax (disabled by default), to be put in markdown as a fenced code blocks.


Also see the README in GitHub Pages.

caption: '*Awesome* **Markdown** Table'
alignment: RC
table-width: 2/3
markdown: True
First row,defaulted to be header row,can be disabled
1,cell can contain **markdown**,"It can be aribrary block element:

- following standard markdown syntax
- like this"
2,"Any markdown syntax, e.g.",E = mc^2^


Awesome Markdown Table
First row defaulted to be header row can be disabled
1 cell can contain markdown

It can be aribrary block element:

  • following standard markdown syntax
  • like this
2 Any markdown syntax, e.g. E = mc2

(The equation might not work if you view this on PyPI.)


pandoc -F pantable -o README.html


Fenced code blocks is used, with a class table. See Example.

Optionally, YAML metadata block can be used within the fenced code block, following standard pandoc YAML metadata block syntax. 7 metadata keys are recognized:


the caption of the table. Can be block-like. If omitted, no caption will be inserted. Interpreted as markdown only if markdown: true below.

Default: disabled.


the short-caption of the table. Must be inline-like element. Interpreted as markdown only if markdown: true below.

Default: disabled.


alignment for columns: a string of characters among L,R,C,D, case-insensitive, corresponds to Left-aligned, Right-aligned, Center-aligned, Default-aligned respectively. e.g. LCRD for a table with 4 columns.

You can specify only the beginning that’s non-default. e.g. DLCR for a table with 8 columns is equivalent to DLCRDDDD.

Default: DDD...


alignment per cell. One row per line. A string of characters among L,R,C,D, case-insensitive, corresponds to Left-aligned, Right-aligned, Center-aligned, Default-aligned respectively. e.g.


for a table with 4 columns, 2 rows.

you can specify only the top left block that is not default, and the rest of the cells with be default to default automatically. e.g.


for a table with 4 columns, 3 rows will be equivalent to


Default: DDD...\n...


a list of relative width corresponding to the width of each columns. D means default width. e.g.

- width
    - 0.1
    - 0.2
    - 0.3
    - 0.4
    - D

Again, you can specify only the left ones that are non-default and it will be padded with defaults.

Default: [D, D, D, ...]


the relative width of the table (e.g. relative to \linewidth). If specified as a number, and if any of the column width in width is default, then auto-width will be performed such that the sum of width equals this number.

Default: None


If it has a header row or not.

Default: True


If CSV table cell contains markdown syntax or not.

Default: False


if true, then the first column of the table will be interpreted as a special fancy-table syntax s.t. it encodes which rows are

  • table-header,
  • table-foot,
  • multiple table-bodies and
  • “body-head” within table-bodies.

see example below.


the path to an CSV file, can be relative/absolute. If non-empty, override the CSV in the CodeBlock.

Default: None

if specified, the file from include will be decoded according to this encoding, else assumed to be UTF-8. Hint: if you save the CSV file via Microsoft Excel, you may need to set this to utf-8-sig.

If specified, should be a dictionary passed to csv.reader as options. e.g.

  dialect: unix
  key: value...

The file format from the data in code-block or include if specified.

Default: csv for data from code-block, and infer from extension in include.

Currently only csv is supported.


(experimental, may drop in the future): a list of int that specifies the number of rows per row-block. e.g. [2, 6, 3, 4, 5, 1] means the table should have 21 rows, first 2 rows are table-head, last 1 row is table-foot, there are 2 table-bodies (indicated by 6, 3, 4, 5 in the middle) where the 1st body 6, 3 has 6 body-head and 3 “body-body”, and the 2nd body 4, 5 has 4 body-head and 5 “body-body”.

If this is specified, header will be ignored.

Default: None, which would be inferred from header.


(experimental, may drop in the future): a list of int that specifies the number of head columns per table-body. e.g. [1, 2] means the 1st table-body has 1 column of head, the 2nd table-body has 2 column of head

Default: None


This one is the inverse of pantable, a panflute filter to convert any native pandoc tables into the CSV table format used by pantable.

Effectively, pantable forms a “CSV Reader”, and pantable2csv forms a “CSV Writer”. It allows you to convert back and forth between these 2 formats.

For example, in the markdown source:

| First  | defaulted to be     | can be disabled          |
| row    | header row          |                          |
| 1      | cell can contain    | It can be aribrary block |
|        | **markdown**        | element:                 |
|        |                     |                          |
|        |                     | -   following standard   |
|        |                     |     markdown syntax      |
|        |                     | -   like this            |
| 2      | Any markdown        | $$E = mc^2$$             |
|        | syntax, e.g.        |                          |

: *Awesome* **Markdown** Table

running pandoc -F pantable2csv -o, it becomes

``` {.table}
alignment: DDD
caption: '*Awesome* **Markdown** Table'
header: true
markdown: true
table-width: 0.8055555555555556
width: [0.125, 0.3055555555555556, 0.375]
First row,defaulted to be header row,can be disabled
1,cell can contain **markdown**,"It can be aribrary block element:

-   following standard markdown syntax
-   like this
2,"Any markdown syntax, e.g.",$$E = mc^2$$


(experimental, may drop in the future)

Similar to pantable2csv, but convert with fancy_table syntax s.t. any general Table in pandoc AST is in principle losslessly converted to a markdown-ish syntax in a CSV representation.


pandoc -F pantable2csvx -o tests/files/native_reference/ tests/files/native/planets.native

would turn the native Table from platnets.native [2] to

``` {.table}
caption: Data about the planets of our solar system.
alignment: CCDRRRRRRRR
- 3
markdown: true
fancy-table: true
===,"(1, 2)
",,Name,Mass (10\^24kg),Diameter (km),Density (kg/m\^3),Gravity (m/s\^2),Length of day (hours),Distance from Sun (10\^6km),Mean temperature (C),Number of moons,Notes
,"(4, 2)
Terrestrial planets",,Mercury,0.330,"4,879",5427,3.7,4222.6,57.9,167,0,Closest to the Sun
,,,Earth,5.97,"12,756",5514,9.8,24.0,149.6,15,1,Our world
,,,Mars,0.642,"6,792",3933,3.7,24.7,227.9,-65,2,The red planet
,"(4, 1)
Jovian planets","(2, 1)
Gas giants",Jupiter,1898,"142,984",1326,23.1,9.9,778.6,-110,67,The largest planet
,,"(2, 1)
Ice giants",Uranus,86.8,"51,118",1271,8.7,17.2,2872.5,-195,27,
___,"(1, 2)
Dwarf planets",,Pluto,0.0146,"2,370",2095,0.7,153.3,5906.4,-225,5,Declassified as a planet in 2006.

Pantable as a library

(experimental, API may change in the future)

Documentation here is sparse, partly because the upstream (pandoc) may change the table AST again. See Crazy ideas: table structure from upstream GitHub.

See the API docs in

For example, looking at the source of pantable as a pandoc filter, in, you will see the main function doing the work is now

pan_table_str = (
    .from_yaml_filter(options=options, data=data, element=element, doc=doc)
if pan_table_str.table_width is not None:
return (

You can see another example from which is what pantable2csv and pantable2csvx called.

Below is a diagram illustrating the API:



Solid arrows are lossless conversions. Dashed arrows are lossy.

You can see the pantable internal structure, PanTable is one-one correspondence to the pandoc Table AST. Similarly for PanCodeBlock.

It can then losslessly converts between PanTable and PanTableMarkdown, where everything in PanTableMarkdown is now markdown strings (whereas those in PanTable are panflute or panflute-like AST objects.)

Lastly, it defines a one-one correspondence to PanCodeBlock with fancy_table syntax mentioned earlier.

Below is the same diagram with the method names. You’d probably want to zoom into it to see it clearly.

Detailed w/ methods

Detailed w/ methods


To run all the tests run tox. GitHub Actions is used for CI too so if you fork this you can check if your commits passes there.

Project details

Download files

Download the file for your platform. If you're not sure which to choose, learn more about installing packages.

Source Distribution

pantable-0.14.2.tar.gz (47.5 kB view hashes)

Uploaded source

Built Distribution

pantable-0.14.2-py3-none-any.whl (44.3 kB view hashes)

Uploaded py3

Supported by

AWS AWS Cloud computing Datadog Datadog Monitoring Facebook / Instagram Facebook / Instagram PSF Sponsor Fastly Fastly CDN Google Google Object Storage and Download Analytics Huawei Huawei PSF Sponsor Microsoft Microsoft PSF Sponsor NVIDIA NVIDIA PSF Sponsor Pingdom Pingdom Monitoring Salesforce Salesforce PSF Sponsor Sentry Sentry Error logging StatusPage StatusPage Status page