lubridate mday() in R: Day of Month Extractor for Dates

The mday() function in lubridate returns the day of the month from a Date or POSIXct value as an integer between 1 and 31. It is the explicit-prefix counterpart to day(), designed to line up visually with wday(), yday(), and qday() in mixed extractor code.

mday(ymd("2024-07-15"))                       # extract day of month
mday(ymd_hms("2024-07-15 09:30:00"))          # works on POSIXct too
mday(c(ymd("2024-01-31"), ymd("2024-06-15"))) # vectorised
mday(x) <- 1                                  # replace day in place
df |> mutate(dom = mday(invoice_date))        # extract a column
mday(x) == days_in_month(x)                   # safe month-end test
mday(x)                                       # same as day(x)
identical(mday(x), day(x))                    # TRUE for any date input

What mday() does in one sentence

mday() returns the day-of-month component of a date-time as an integer. Pass any Date, POSIXct, or POSIXlt vector and you get a numeric vector of the same length with values 1 through 31.

The "m" prefix stands for "month". It is the lubridate counterpart to base R's as.POSIXlt(x)$mday, but returns an integer you can use directly in arithmetic and comparisons.

Syntax

mday(x) accepts a single argument: a date-time vector. It returns an integer vector with values 1 to 31. Unlike wday(), there are no label or week_start arguments because day-of-month numbers are unambiguous.

The input must be a Date, POSIXct, or POSIXlt value. Character strings are NOT accepted; parse them first with ymd(), mdy(), or as.Date().

Six common patterns

1. Extract day of month from a single date

The result is integer 15, not the string "15". Use it in arithmetic directly: mday(d) + 7 returns 22.

2. Extract day from a vector of dates

mday() is fully vectorised: a million-row date column becomes a million-row integer vector in one call. The result respects each month's length, so February 29 in a leap year returns 29 cleanly.

3. Pair mday() with wday(), yday(), and qday() in one pipeline

This is the feature-engineering use case: four day-position numbers from one date column. mday() reads cleanly alongside wday(), yday(), and qday() because all four share a one-letter calendar prefix.

4. Replace mday values in place

The replacement form keeps year, month, and time-of-day untouched. This buckets timestamps to the start of their month without a floor_date(x, "month") call.

5. Filter rows by day of month

mday(invoice_date) >= 25 keeps dates in the last week of any month, regardless of month length. Combine with between() for windows: filter(between(mday(invoice_date), 10, 20)) keeps mid-month rows.

6. Safe month-end test with days_in_month()

days_in_month() returns the last valid day for each month (28, 29, 30, or 31) and respects leap years. mday(x) == days_in_month(x) is the only correct month-end flag; hard-coding mday(x) == 31 silently drops every February, every 30-day month, and breaks on leap years.

mday() vs day(): when to write which

mday() and day() are exact aliases; both return the day-of-month integer. The choice is a readability concern, not a behavioural one.

Verify the equivalence in one line:

Both lubridate functions agree with the base R format(x, "%d") string, cast to integer. There is no version, time zone, or class on which mday() and day() diverge.

Common pitfalls

Pitfall 1: passing a character string. mday("2024-07-15") errors because lubridate does not auto-parse strings inside extractors. Wrap with ymd() first: mday(ymd("2024-07-15")). The same applies to slash-separated and US-style strings; parse, then extract.

Pitfall 2: assigning 31 to a 30-day month. mday(ymd("2024-04-15")) <- 31 returns "2024-05-01" because April has 30 days and lubridate rolls forward instead of erroring. Cap with pmin(target_day, days_in_month(x)) for fail-fast behaviour.

Pitfall 3: hard-coding 31 as "the last day". mday(x) == 31 drops every February row and every 30-day month-end. Pair with days_in_month(x) instead.

A practical workflow with mday()

Day-of-month shows up in three places: feature engineering, window filters, and safe end-of-month checks.

The patterns:

1. Feature engineering for models. Pair mday() with wday(), yday(), and qday() so the four columns read as parallel calendar positions. Use the three-letter form for all four to keep the code grid-aligned.
2. Within-month windows. filter(mday(date) <= 7) keeps the first week of every month; filter(mday(date) >= 25) keeps the last week. Combine with month(date) if you need a specific month and the last week.
3. Robust month-end logic. mday(x) == days_in_month(x) is the only filter that respects February, 30-day months, and leap years. Use it instead of mday(x) == 31.

For charts, prefer floor_date(x, "day") over mday() when you need a real date axis. mday() collapses three years of January 15th into the integer 15, losing year and month context.

Aggregating by mday() collapses a year of sales into 31 buckets, one per calendar day. The result answers "are payday days higher on average?" without committing to a specific month or year.

Try it yourself

Related lubridate functions

After mastering mday(), look at:

• day(): identical alias for day of month; use whichever reads best
• wday(), yday(), qday(): the rest of the day-prefix family
• year(), month(): extract the other calendar parts
• hour(), minute(), second(): extract time-of-day components
• days_in_month(): get the last valid day for any month
• floor_date(), ceiling_date(): round a date to day, week, month, or year
• make_date(), make_datetime(): build a date from year, month, day integers

For the official reference, see the lubridate day() documentation, which covers mday() alongside the rest of the family.

FAQ

What does mday() return in R?

mday() returns the day of the month as an integer between 1 and 31. Input is a Date, POSIXct, or POSIXlt vector; output is an integer vector of the same length. The function is fully vectorised, so it works on a single date or a million-row time-series without a loop. Character strings are not accepted; parse them with ymd(), mdy(), or as.Date() first.

Is mday() the same as day() in lubridate?

Yes, mday() and day() are exact aliases. Both return the day-of-month integer and identical(mday(x), day(x)) is TRUE for any valid input. The two names exist for readability: mday() lines up with wday(), yday(), and qday() in mixed-extractor code, while day() reads cleaner standalone.

How do I get the last day of the month in R?

Use days_in_month(x) to get the integer 28, 29, 30, or 31 for any date, then compare with mday(x). For example, mday(x) == days_in_month(x) flags true month-end rows and correctly handles February, 30-day months, and leap years. Hard-coding mday(x) == 31 silently drops every February row and every short-month month-end.

Can I change the day of a date with mday()?

Yes, use the replacement form: mday(x) <- 1. This sets the day while keeping year, month, and time-of-day untouched. The form is vectorised, so mday(date_column) <- 1 snaps every row to the first of its month. Values above the month's length roll forward (mday(ymd("2024-04-15")) <- 31 becomes May 1), so cap with pmin(value, days_in_month(x)) for fail-fast behaviour.

What is the difference between mday() and wday() in R?

mday() returns the position of a date within its month (1 to 31); wday() returns the position within the week (1 to 7, with Sunday as 1 by default). Both return small integers, but they answer different questions. Use mday() for calendar-day analysis (paydays, billing cycles) and wday(x, label = TRUE) for weekday analysis.