Fast, dependency-free terminal tables for Python.
54 border styles · ANSI colours · cell merges · decimal alignment · width fitting · CSV / Markdown / HTML / Excel / LaTeX I/O

flexablelib turns rows and columns into readable tables for terminals, logs, reports and docs. It has no runtime dependencies, ships prebuilt wheels with an optional C speed-up, and is covered by 1,000+ tests on Python 3.8 – 3.14.

Install

pip install flexablelib

Optional extras: pip install "flexablelib[pandas]", "flexablelib[excel]", "flexablelib[html]" or "flexablelib[all]". The base install stays dependency-free.

30-second quickstart

from flexablelib import Table

table = Table(
    rows=[
        ["api-gateway", "eu-west-1", 99.982, 1240],
        ["auth", "us-east-1", 99.9, 312],
        ["billing", "us-east-1", 98.75, 87],
    ],
    headers=["Service", "Region", "Uptime %", "RPS"],
)
table.style_name = "rounded"            # 54 styles, see style_examples.md
table.header_color = "bold cyan"
table.color_rule = lambda value, row, column: (
    "red" if column == "Uptime %" and value < 99.5 else None
)
print(table)

╭─────────────┬───────────┬──────────┬──────╮
│ Service     │ Region    │ Uptime % │  RPS │
├─────────────┼───────────┼──────────┼──────┤
│ api-gateway │ eu-west-1 │   99.982 │ 1240 │
│ auth        │ us-east-1 │   99.9   │  312 │
│ billing     │ us-east-1 │   98.75  │   87 │
╰─────────────┴───────────┴──────────┴──────╯

Numbers align on the decimal point, colours never break alignment, and the same table exports anywhere:

table.to_markdown()                # GitHub / docs
table.to_html("report.html")       # self-contained, sortable, offline
table.to_csv("out.csv"); table.to_latex(); table.to_rst(); table.to_asciidoc()
table.to_excel("report.xlsx")      # pip install "flexablelib[excel]"
Table.from_csv("sales.csv"); Table.from_markdown(md); Table.from_pandas(df)

Feature gallery

Styles — 54 presets, from ascii_rows to borderless	Themes — table.apply_theme("ocean")
Text on borders — table.border_text("Production", where="top")	Cell merges — table.merge_cells(0, 0, last_row=1)
Zebra rows, footer, caption	Decimal alignment — floats and scientific notation

More: all 54 styles · wrapping and ellipsis overflow · per-column / per-cell colour and alignment · vertical alignment · paginate() · stream() for live rows · transpose() · nested tables · fit_terminal() · hide_columns · sort_by · Roman / custom index.

How it compares

A factual feature matrix from our survey of 89 table libraries. A dash means the feature was not found in that library's documentation during the survey, not a claim that it is impossible there.

rich is a full terminal-rendering toolkit (markup, live displays, progress bars) and tabulate covers many more markup dialects; pick flexablelib when you want a dependency-free, table-focused library with styling, merges and I/O in one place.

Production notes

• Zero runtime dependencies. Readers/writers that need a third-party package (pandas, openpyxl, lxml) are opt-in extras and imported lazily.
• Performance. Rendering 10,000 rows × 6 columns takes ~157 ms with the C extension (~172 ms pure Python), on par with prettytable, ~1.3× faster than tabulate and ~18× faster than rich's table. pandas to_string() is faster still (~85 ms). Full method and numbers: docs/BENCHMARKS.md.
• Wheels. Prebuilt for CPython 3.8 – 3.14 on Linux (manylinux, musllinux), Windows x64 and macOS arm64. Anywhere else it builds from source or falls back to pure Python — installation never fails for want of a compiler. from flexablelib.fast import implementation; implementation() tells you which backend is live.
• Colour safety. Escape sequences are emitted only to a TTY by default, never when NO_COLOR is set; FORCE_COLOR or table.use_colors = True overrides. Logs and files stay clean.
• Unicode. Widths are measured in terminal columns, so CJK, emoji and coloured text stay aligned.
• Windows. Box-drawing styles need a UTF-8 console (chcp 65001 or PYTHONUTF8=1; Windows Terminal is usually UTF-8 already). ASCII styles (ascii_rows, ascii_double_underline_rows, psql, markdown, …) work on any code page.
• Typed. Ships py.typed (PEP 561).
• Stability. flexablelib follows semantic versioning: within 2.x, public Table attributes and methods are not removed or renamed without a deprecation period; new features arrive in minor/patch releases and every release is listed in the CHANGELOG.
• Support. Bugs and requests: issues. Security reports: SECURITY.md. Contributions: CONTRIBUTING.md.

Windows console encoding

chcp 65001
set PYTHONUTF8=1

(PowerShell: $env:PYTHONUTF8=1.)

Guide

Usage

Creating a table is simple.

Code Example

from flexablelib import Table

new_table = Table()
print(new_table)

Output

++
++
++

This is an empty table. It has no data so it only displays this strange thing.

A table that has columns but no rows — a query that came back empty — shows its headers rather than raising:

┌──────┬───────┐
│ name │ value │
╞══════╪═══════╡
└──────┴───────┘

It's possible to add data as columns or rows, or even alternating each one. Any untitled column will be named automatically.

Code Example

new_table.add_column('Name', ['Jade', 'John', 'Jane'])
new_table.add_column('Age', [20, 30, 40])
new_table.add_column('Results', [9.651, 3, 245.7])
print(new_table)

Output

+----------------------+
| Name   Age   Results |
+======+=====+=========+
| Jade |  20 |   9.651 |
| John |  30 |   3     |
| Jane |  40 | 245.7   |
+------+-----+---------+

Code Example

new_table = Table()
new_table.add_column('Name', ['Jade', 'John', 'Jane'])
new_table.add_column('Age', [20, 30, 40])
new_table.add_column('Test\nResults', [9.651, 3, 245.7])
new_table.add_row(['Piotr\nBaltimore', 27, 3.5])
new_table.add_row(['Sam', 21, 0.6519])
print(new_table)

Output

+----------------------------+
| Name        Age       Test |
|                    Results |
+===========+=====+==========+
| Jade      |  20 |   9.651  |
| John      |  30 |   3      |
| Jane      |  40 | 245.7    |
| Piotr     |  27 |   3.5    |
| Baltimore |     |          |
| Sam       |  21 |   0.6519 |
+-----------+-----+----------+

As it is visible, the table will format automatically new lines and data types, for now without trying to parse strings that could be converted to another type.

The Table class offers a variety of options that allow things like showing the index of each row, changing the style of the table, hiding the headers, etc.

See style examples here.

Code Example

new_table = Table()
new_table.add_column('Name', ['Jade', 'John'])
new_table.add_column('Age', [20, 30])
new_table.add_column('Test\nResults', [9.651, 3, 245.7])
new_table.add_row(['Piotr\nBaltimore', 27, 3.5])
new_table.add_row(['Sam', 21])
new_table.show_index = True
new_table.style_name = 'double_horizontal'
new_table.missing_value = '?'
print(new_table)

Output

╒═══╤═══════════╤═════╤═════════╕
│ i │ Name      │ Age │    Test │
│   │           │     │ Results │
╞═══╪═══════════╪═════╪═════════╡
│ 0 │ Jade      │  20 │   9.651 │
│ 1 │ John      │  30 │   3     │
│ 2 │ ?         │   ? │ 245.7   │
│ 3 │ Piotr     │  27 │   3.5   │
│   │ Baltimore │     │         │
│ 4 │ Sam       │  21 │       ? │
╘═══╧═══════════╧═════╧═════════╛

The missing value aligns as if it was of the same type of the other data in the column.

Changed in 2.3.0: an explicit None is a missing cell, just like a padding gap or table.missing. It displays missing_value (blank by default), not the text "None". Set table.missing_value = '?' to show gaps explicitly. Use the string "None" if you want that literal text. This applies to printing, compose(), pagination, streaming, and transposition in every style.

Missing cells do not turn numeric columns into text: numeric alignment and the decimal axis are preserved, and overflow='wrap' does not split floats because a column contains None. Raw rows and exports still represent gaps as None; rendering Table(rows=table.rows) reproduces the same cells (copy headers and display settings too if you changed them).

table = Table(rows=[[924.87], [None], [1.25]])
assert 'None' not in table.compose()
assert Table(rows=table.rows).compose() == table.compose()
table.missing_value = '?'  # changes the display, not stored None values

See missing-cell semantics for migration details.

This shows how to get the row and column count. If the index is shown, this count remains unaffected by that column, although, you can get the internal count.

Code Example

new_table.show_index = False
new_table.show_headers = False
print(new_table)
print('row:', new_table.row_count)
print('columns:', new_table.column_count)
new_table.show_index = True
print(new_table)
print('internal_row_count:', new_table.internal_row_count)
print('internal_column_count:', new_table.internal_column_count)

Output

╒═══════════╤════╤═════════╕
│ Jade      │ 20 │   9.651 │
│ John      │ 30 │   3     │
│ ?         │  ? │ 245.7   │
│ Piotr     │ 27 │   3.5   │
│ Baltimore │    │         │
│ Sam       │ 21 │       ? │
╘═══════════╧════╧═════════╛
row: 5
columns: 3
╒═══╤═══════════╤════╤═════════╕
│ 0 │ Jade      │ 20 │   9.651 │
│ 1 │ John      │ 30 │   3     │
│ 2 │ ?         │  ? │ 245.7   │
│ 3 │ Piotr     │ 27 │   3.5   │
│   │ Baltimore │    │         │
│ 4 │ Sam       │ 21 │       ? │
╘═══╧═══════════╧════╧═════════╛
internal_row_count: 5
internal_column_count: 4

Colour

Colour the header, the borders, whole columns, whole rows, or individual cells. Specs accept names, attributes, 256-colour indexes and hex triples.

table.header_color = 'bold cyan'
table.border_color = 'grey'
table.column_colors = {'Service': 'bright_white'}
table.color_rule = lambda value, row, column: (
    'red' if column == 'Delta' and value < 0 else None
)

color_rule receives the original value, so numeric comparisons work directly.

Colour is emitted only to a terminal by default, and never when NO_COLOR is set. FORCE_COLOR or table.use_colors = True overrides that. Because compose() returns a string you may send anywhere, the default errs toward not embedding escape sequences in something bound for a file.

Widths are measured in terminal columns, not characters, so coloured cells and CJK or emoji data stay aligned.

Reading and writing

Table.from_csv('sales.csv')          # numeric columns parsed and aligned
Table.from_html(markup)              # standard library parser, no dependency
Table.from_dicts(records)
Table.from_pandas(dataframe)         # pip install flexablelib[pandas]
Table.from_excel('report.xlsx')      # pip install flexablelib[excel]

table.to_csv('out.csv')
table.to_markdown()
table.to_html('report.html', paginate=25)
table.to_excel('report.xlsx')

Since 2.2.0 every export writes the table as print(table) shows it: rows in sort_by order (with sort_reverse and callables), row_filter applied, hide_columns left out, and the text formats (to_csv, to_markdown, to_html, to_rst, to_asciidoc, to_latex) hold the formatted strings — float_format, int_format, leading_zeros, custom_format, strip_cells — without padding or colour. Empty cells are empty strings.

table.sort_by, table.sort_reverse = 'Uptime', True
table.row_filter = lambda row: row[1] < 99.9
table.float_format = '.1f'
table.to_markdown()                  # same rows, order and numbers as print(table)

to_records() keeps the stored Python values (numbers stay numbers, gaps are None) in display order; to_records(formatted=True) gives the printed strings. to_excel() writes numbers as numbers and turns simple format specs (',d', '.2f', '.1%', '{:,.2f}'…) into the cell's Excel number format.

to_html writes one self-contained file — the stylesheet and the sorting, filtering and pagination script are inline — so it works offline and from a file:// URL.

Text formats deliver everything as strings, and a column of strings is left-aligned. parse_str_numbers converts numeric-looking text so it aligns as numbers; the readers turn it on for you:

table.parse_str_numbers = True

Values with leading zeros are left alone, so '007' stays a string.

table.rows is a copy of the stored data, in insertion order and with None in empty cells (sorting, filtering and formats do not touch it), so editing it does not change the table. Use add_row and set_cell instead:

table.set_cell(0, 'Uptime', 98.2)    # row position, column position or header
table.set_cell(1, 2, table.missing)  # empty a cell

Merging cells

Render a rectangular block as a single cell, across columns, down rows, or both. Coordinates are inclusive, zero-based, and ignore the index column.

table.merge_cells(0, 0, last_column=2, value='Quarter summary')
table.merge_cells(1, 0, last_row=3)                # keeps the top-left content
table.merge_cells(0, 0, 2, 2, value='Total', align='r')

A merge never widens the table -- columns are still sized by their unmerged content -- so text longer than its span is truncated.

Notes

• A column named "i" does not collide with the index column (utils.IndexColumnTitle).
• Exponential numbers align on the decimal / exponent axis with the rest of the float column.
• Shrinking a float column drops decimals instead of chopping characters, so decimal alignment survives a terminal fit (#23).

Open work and the issue map: docs/ISSUES.md.

Project Layout

flexablelib/                 # GitHub repository
├── flexablelib/            the Python package (no runtime dependencies)
│   ├── __init__.py          public API: Table, TableComposition, SeparatorLine
│   ├── table.py             the Table class — state and render orchestration
│   ├── columns.py           type inference, alignment, column widths
│   ├── style_compositions.py   the 54 border styles, as data
│   ├── table_strings.py     separator lines and data rows
│   ├── cells.py             cell padding, justification, wrapping
│   ├── options.py           constants and defaults
│   └── utils.py             type predicates and small helpers
├── tests/                   pytest suite — see ARCHITECTURE.md#testing
├── docs/                    issue history and design notes
├── logos/                   brand assets — see logos/README.md
├── ARCHITECTURE.md          how it all fits together
├── style_examples.md        all 54 styles rendered
├── package.json             canonical version number (no npm deps)
└── setup.py                 reads the version from package.json

Run the tests from the repository root:

pip install -r requirements-dev.txt
python -m pytest

See ARCHITECTURE.md for the rendering pipeline, how styles are defined, and how to add one.