busdayaxis

A Matplotlib scale that compresses non-business days and off-hours. Every visible unit on the axis corresponds to active time — no gaps for weekends, holidays, or overnight periods. No data preprocessing needed.

Full documentation at saemeon.github.io/busdayaxis

Installation

pip install busdayaxis

Quick Start

import matplotlib.pyplot as plt
import pandas as pd
import busdayaxis  # scale is registered automatically on import

dates = pd.date_range("2026-01-01", periods=10, freq="B")
values = range(len(dates))

fig, ax = plt.subplots()
ax.plot(dates, values)
ax.set_xscale("busday", bushours=(9, 17))  # compress weekends (Mon–Fri default) and off-hours (business hours: 09:00–17:00)
plt.show()

Why

Time series that only evolve on business days — prices, signals, operational metrics — look distorted on a standard calendar axis. Weekends and holidays introduce flat gaps that compress active periods and visually skew slopes. busdayaxis removes these gaps entirely.

What it provides

• Compress weekends, holidays, and overnight gaps by adding an axis scalewith a single call:

  ax.set_xscale("busday", weekmask=..., holidays=..., busdaycal=..., bushours=...)
  

  • weekmask, holidays, busdaycal: standard numpy.is_busday parameters to configure which days are considered business days

  • bushours: define uniform or weekday-specific business hours

  • Implemented as a proper matplotlib.scale — autoscaling, shared axes, and all standard artists work without any changes to your plotting code

• Business-day-aware DateLocator wrappers for all standard matplotlib.dates locators — automatically filter out ticks on off-days and off-hours

• BusdayLocator base class to wrap any custom or third-party date locator with the same business-day filtering logic

• MidBusdayLocator to place a tick at the midpoint of each business session, useful for centering day labels

• mark_gaps to draw visual indicators (vlines or broken-axis slash marks) at every session boundary on a busday axis

• holidays_from_exchange to build the holidays= list automatically from any exchange_calendars or pandas_market_calendars calendar object

Under the Hood

• matplotlib internally handles dates as floating-point numbers representing days since 1970-01-01, or stated alternatively, as

  $$\text{ matplotlib-representation}=\frac{\text{hours\ since\ 1970-01-01}}{24\ \text{hours}}$$

• busdayaxis transforms these coordinates to floating-point numbers representing

  $$\text{busdayaxis-representation} = \frac{\text{business-hours since 1970-01-01}}{24 \text{ hours}}$$

  This conversion implies that datetime values that fall on non-business days or outside of business hours will be mapped to the same coordinate as the nearest preceding business hour. For example, if business hours are defined as 9:00 to 17:00, then "1970-01-05 08:00" (Mon 08:00) is mapped to the same coordinate as "1970-01-05 09:00" (Mon 09:00), because the earlier timestamp lies outside the defined business hours.

Example: "1970-01-05 10:00" (Mon 10:00)

• Matplotlib representation (all hours counted):

      Thu 1970-01-01   24h (00:00 - 24:00)
    + Fri 1970-01-02   24h (00:00 - 24:00)
    + Sat 1970-01-03   24h (00:00 - 24:00)
    + Sun 1970-01-04   24h (00:00 - 24:00)
    + Mon 1970-01-05   10h (00:00 - 10:00)
    ---------------------------------------
    =                 106h (Total hours since epoch)
    /                  24h
    ---------------------------------------
    =                  4.41666... (matplotlib coordinate)
  

• Busdayaxis representation:

  • We assume here that business hours are from 9:00 to 17:00. This can be configured by setting ax.set_xscale("busday", bushours=(9, 17)).

  To get the floating-point representation of "1970-01-05 10:00" (Mon 10:00), we count the business hours that have elapsed since 1970-01-01 00:00:

      Thu 1970-01-01   8h (9:00 - 17:00)
    + Fri 1970-01-02   8h (9:00 - 17:00)
    + Sat 1970-01-03   0h
    + Sun 1970-01-04   0h
    + Mon 1970-01-05   1h (09:00 - 10:00)
    ---------------------------------------
    =                 17h (business hours since epoch)
    /                 24h
    ---------------------------------------
    =          0.708333... (busdayaxis coordinate)
  

License

MIT