pandas_market_calendars

What it is

A pandas-shaped answer to "when was this market open". mcal.get_calendar("NYSE") returns a calendar, and schedule(start_date, end_date) returns a DataFrame with one row per session and market_open and market_close columns in UTC — plus break_start and break_end where a market pauses, and pre and post when you pass start="pre", end="post". mcal.date_range(schedule, frequency="1h") turns that into a DatetimeIndex of bar timestamps that skips closed hours and cuts the last bar at an early close. merge_schedules intersects or unions several schedules, open_at_time tests a timestamp, and valid_days, early_closes and interruptions_df answer the narrower questions.

Ryan Sheftel started it in December 2016 as a fork of the calendar code in Quantopian's Zipline. Since version 2.0 it also mirrors every calendar in exchange_calendars and depends on that package, so the choice between the two is mostly a choice of API.

Pricing

Free under the MIT licence; nothing to buy. pip install pandas_market_calendars, Python 3.10 or later, with pandas and exchange_calendars as its only declared dependencies. An Arch Linux AUR package also exists.

Data & coverage

What this package adds over the mirror is futures, options and fixed income. Version 5.4.0 lists 211 calendar names including aliases, 106 distinct calendars: 70 mirrored from exchange_calendars and 36 of its own. The 36 cover fourteen equity exchanges written natively (NYSE, LSE, TSX, B3, JPX, OSE, HKEX, SSE, NSE, BSE, ASX, SIX, TASE, IEX), CME product calendars by Globex group — equity, FX, fixed income, energy and metals, grains, livestock, crypto — plus a CME_TradeDate calendar for settlement-date users, ICE and the CBOE Futures Exchange, separate CBOE equity-options and index-options calendars (index options close at 18:15 UTC on the day after Thanksgiving 2026, equity options at 18:00), Eurex with a pre- and post-session variant, SIFMA bond-market calendars for the US, UK and Japan, and a forex calendar.

CME calendars use Globex session dates: the 23 December 2026 CME_Equity session opens at 23:00 UTC on the 22nd. Since 5.3.0 a calendar can carry sources — the NYSE class cites the exchange's hours page, last verified 24 January 2025.

The calendars are package code, not a feed. Nothing is fetched at runtime, and a correction reaches you only by upgrading.

Integrations

Built for pandas: schedules are DataFrames, ranges are DatetimeIndexes, and time zones use the standard library's zoneinfo since version 5.0. exchange_calendars sits underneath, so a reader already using XNYS-style codes keeps them.

Limitations

• Fixes reach dev before PyPI. Issue #464 reported five wrong 2026 early closes on CME Globex energy and metals; the fix was merged to dev on 12 July 2026 and is in no release. master last moved on 27 May 2026.
• Two layers to keep current. The mirrored calendars are whatever exchange_calendars version is installed, and that project has its own release lag.
• Two calendars for one exchange. NYSE and XNYS are separate code paths; agreeing over two decades today is no promise both receive tomorrow's fix at once.
• Contributed calendars. The docs credit individual creators per calendar, and several have no unit tests on the status page — TASE, ASX and BMF among them.
• Regular sessions only by default; pre- and post-market need explicit arguments and exist only where a calendar defines them.

Alternatives

exchange_calendars on its own, if equity sessions under ISO codes are all you need — one dependency fewer. A trading calendar as data sets both libraries beside the market-holiday endpoints of data APIs, and covers when a trading calendar is the wrong tool for counting settlement days.