parse_date_time: User friendly date-time parsing functions

Description

parse_date_time() parses an input vector into POSIXct date-time object. It differs from base::strptime() in two respects. First, it allows specification of the order in which the formats occur without the need to include separators and % prefix. Such a formating argument is refered to as "order". Second, it allows the user to specify several format-orders to handle heterogeneous date-time character representations.

parse_date_time2() is a fast C parser of numeric orders.

fast_strptime() is a fast C parser of numeric formats only that accepts explicit format arguments, just as base::strptime().

Usage

parse_date_time(x, orders, tz = "UTC", truncated = 0, quiet = FALSE,
  locale = Sys.getlocale("LC_TIME"), select_formats = .select_formats,
  exact = FALSE, train = TRUE, drop = FALSE)
parse_date_time2(x, orders, tz = "UTC", exact = FALSE, lt = FALSE,
  cutoff_2000 = 68L)
fast_strptime(x, format, tz = "UTC", lt = TRUE, cutoff_2000 = 68L)

Arguments

a character or numeric vector of dates

orders

a character vector of date-time formats. Each order string is series of formatting characters as listed base::strptime() but might not include the "%" prefix, for example "ymd" will match all the possible dates in year, month, day order. Formatting orders might include arbitrary separators. These are discarded. See details for implemented formats.

a character string that specifies the time zone with which to parse the dates

truncated

integer, number of formats that can be missing. The most common type of irregularity in date-time data is the truncation due to rounding or unavailability of the time stamp. If the truncated parameter is non-zero parse_date_time() also checks for truncated formats. For example, if the format order is "ymdHMS" and truncated = 3, parse_date_time() will correctly parse incomplete dates like 2012-06-01 12:23, 2012-06-01 12 and 2012-06-01. NOTE: The ymd family of functions are based on strptime() which currently fails to parse %Y-%m formats.

quiet

logical. When TRUE progress messages are not printed, and "no formats found" error is surpresed and the function simply returns a vector of NAs. This mirrors the behavior of base R functions strptime() and as.POSIXct(). Default is FALSE.

locale

locale to be used, see locales. On linux systems you can use system("locale -a") to list all the installed locales.

select_formats

A function to select actual formats for parsing from a set of formats which matched a training subset of x. It receives a named integer vector and returns a character vector of selected formats. Names of the input vector are formats (not orders) that matched the training set. Numeric values are the number of dates (in the training set) that matched the corresponding format. You should use this argument if the default selection method fails to select the formats in the right order. By default the formats with most formating tokens (%) are selected and %Y counts as 2.5 tokens (so that it has a priority over %y%m). See examples.

exact

logical. If TRUE, the orders parameter is interpreted as an exact strptime() format and no training or guessing are performed (i.e. train, drop parameters are irrelevant).

train

logical, default TRUE. Whether to train formats on a subset of the input vector. The result of this is that supplied orders are sorted according to performance on this training set, which commonly results in increased performance. Please note that even when train is FALSE (and exact is FALSE) guessing of the actual formats is still performed on a pseudo-random subset of the original input vector. This might result in All formats failed to parse error. See notes below.

drop

logical, default FALSE. Whether to drop formats that didn't match on the training set. If FALSE, unmatched on the training set formats are tried as a last resort at the end of the parsing queue. Applies only when train=TRUE. Seating this parameter to TRUE might slightly speed up parsing in situations involving many formats. Prior to v1.7.0 this parameter was implicitly TRUE, which resulted in occasional surprising behavior when rare patterns where not present in the training set.

logical. If TRUE returned object is of class POSIXlt, and POSIXct otherwise. For compatibility with base strptime function default is TRUE for fast_strptime and FALSE for parse_date_time2.

cutoff_2000

integer. For y format, two-digit numbers smaller or equal to cutoff_2000 are parsed as 20th's century, 19th's otherwise. Available only for functions relying on lubridates internal parser.

format

a character string of formats. It should include all the separators and each format must be prefixed with argument of strptime().

Value

a vector of POSIXct date-time objects

Details

When several format-orders are specified, parse_date_time() selects (guesses) format-orders based on a training sub-set of the input strings. After guessing the formats are ordered according to the performance on the training set and applied recursively on the entire input vector. You can disable training with train=FALSE.

parse_date_time(), and all derived functions, such as ymd_hms(), ymd() etc, will drop into fast_strptime() instead of strptime() whenever the guessed from the input data formats are all numeric.

The list below contains formats recognized by lubridate. For numeric formats leading 0s are optional. As compared to base strptime(), some of the formats are new or have been extended for efficiency reasons. These formats are marked with "*". The fast parsers parse_date_time2() and fast_strptime() accept only formats marked with "!".

a

Abbreviated weekday name in the current locale. (Also matches full name)

A

Full weekday name in the current locale. (Also matches abbreviated name).

You don't need to specify a and A formats explicitly. Wday is automatically handled if preproc_wday = TRUE

b!

Abbreviated or full month name in the current locale. The C parser currently understands only English month names.

B!

Same as b.

d!

Day of the month as decimal number (01--31 or 0--31)

H!

Hours as decimal number (00--24 or 0--24).

I!

Hours as decimal number (01--12 or 1--12).

j

Day of year as decimal number (001--366 or 1--366).

q!*

Quarter (1--4). The quarter month is added to the parsed month if m format is present.

m!*

Month as decimal number (01--12 or 1--12). For parse_date_time. As a lubridate extension, also matches abbreviated and full months names as b and B formats. C parser understands only English month names.

M!

Minute as decimal number (00--59 or 0--59).

p!

AM/PM indicator in the locale. Normally used in conjunction with I and not with H. But the lubridate C parser accepts H format as long as hour is not greater than 12. C parser understands only English locale AM/PM indicator.

S!

Second as decimal number (00--61 or 0--61), allowing for up to two leap-seconds (but POSIX-compliant implementations will ignore leap seconds).

OS

Fractional second.

U

Week of the year as decimal number (00--53 or 0--53) using Sunday as the first day 1 of the week (and typically with the first Sunday of the year as day 1 of week 1). The US convention.

w

Weekday as decimal number (0--6, Sunday is 0).

W

Week of the year as decimal number (00--53 or 0--53) using Monday as the first day of week (and typically with the first Monday of the year as day 1 of week 1). The UK convention.

y!*

Year without century (00--99 or 0--99). In parse_date_time() also matches year with century (Y format).

Y!

Year with century.

z!*

ISO8601 signed offset in hours and minutes from UTC. For example -0800, -08:00 or -08, all represent 8 hours behind UTC. This format also matches the Z (Zulu) UTC indicator. Because strptime doesn't fully support ISO8601 this format is implemented as an union of 4 orders: Ou (Z), Oz (-0800), OO (-08:00) and Oo (-08). You can use these four orders as any other but it is rarely necessary. parse_date_time2() and fast_strptime() support all of the timezone formats.

Om!*

Matches numeric month and English alphabetic months (Both, long and abbreviated forms).

Op!*

Matches AM/PM English indicator.

r*

Matches Ip and H orders.

R*

Matches HM andIMp orders.

T*

Matches IMSp, HMS, and HMOS orders.

Examples

Run this code

# NOT RUN {
## ** orders are much easier to write **
x <- c("09-01-01", "09-01-02", "09-01-03")
parse_date_time(x, "ymd")
parse_date_time(x, "y m d")
parse_date_time(x, "%y%m%d")
#  "2009-01-01 UTC" "2009-01-02 UTC" "2009-01-03 UTC"

## ** heterogenuous date-times **
x <- c("09-01-01", "090102", "09-01 03", "09-01-03 12:02")
parse_date_time(x, c("ymd", "ymd HM"))

## ** different ymd orders **
x <- c("2009-01-01", "02022010", "02-02-2010")
parse_date_time(x, c("dmY", "ymd"))
##  "2009-01-01 UTC" "2010-02-02 UTC" "2010-02-02 UTC"

## ** truncated time-dates **
x <- c("2011-12-31 12:59:59", "2010-01-01 12:11", "2010-01-01 12", "2010-01-01")
parse_date_time(x, "Ymd HMS", truncated = 3)

## ** specifying exact formats and avoiding training and guessing **
parse_date_time(x, c("%m-%d-%y", "%m%d%y", "%m-%d-%y %H:%M"), exact = TRUE)
parse_date_time(c('12/17/1996 04:00:00','4/18/1950 0130'),
                c('%m/%d/%Y %I:%M:%S','%m/%d/%Y %H%M'), exact = TRUE)

## ** quarters and partial dates **
parse_date_time(c("2016.2", "2016-04"), orders = "Yq")
parse_date_time(c("2016", "2016-04"), orders = c("Y", "Ym"))

## ** fast parsing **
# }
# NOT RUN {
  options(digits.secs = 3)
  ## random times between 1400 and 3000
  tt <- as.character(.POSIXct(runif(1000, -17987443200, 32503680000)))
  tt <- rep.int(tt, 1000)

  system.time(out <- as.POSIXct(tt, tz = "UTC"))
  system.time(out1 <- ymd_hms(tt)) # constant overhead on long vectors
  system.time(out2 <- parse_date_time2(tt, "YmdHMOS"))
  system.time(out3 <- fast_strptime(tt, "%Y-%m-%d %H:%M:%OS"))

  all.equal(out, out1)
  all.equal(out, out2)
  all.equal(out, out3)
# }
# NOT RUN {
## ** how to use `select_formats` argument **
## By default %Y has precedence:
parse_date_time(c("27-09-13", "27-09-2013"), "dmy")

## to give priority to %y format, define your own select_format function:

my_select <-   function(trained, drop=FALSE, ...){
   n_fmts <- nchar(gsub("[^%]", "", names(trained))) + grepl("%y", names(trained))*1.5
   names(trained[ which.max(n_fmts) ])
}

parse_date_time(c("27-09-13", "27-09-2013"), "dmy", select_formats = my_select)

## ** invalid times with "fast" parsing **
parse_date_time("2010-03-14 02:05:06",  "YmdHMS", tz = "America/New_York")
parse_date_time2("2010-03-14 02:05:06",  "YmdHMS", tz = "America/New_York")
parse_date_time2("2010-03-14 02:05:06",  "YmdHMS", tz = "America/New_York", lt = TRUE)
# }