Learn R Programming

xts (version 0.14.1)

index.xts: Get and Replace the Class of an xts Index

Description

Functions to get and replace an xts object's index values and it's components.

Usage

# S3 method for xts
index(x, ...)

# S3 method for xts index(x) <- value

# S3 method for xts time(x) <- value

# S3 method for xts time(x, ...)

.index(x, ...)

.index(x) <- value

.indexsec(x)

.indexmin(x)

.indexhour(x)

.indexmday(x)

.indexmon(x)

.indexyear(x)

.indexwday(x)

.indexbday(x)

.indexyday(x)

.indexisdst(x)

.indexDate(x)

.indexday(x)

.indexweek(x)

.indexyweek(x)

convertIndex(x, value)

Arguments

x

An xts object.

...

Arguments passed to other methods.

value

A new time index value.

Author

Jeffrey A. Ryan

Details

An xts object's index is stored internally as the number of seconds since UNIX epoch in the UTC timezone. The .index() and .index<- functions get and replace the internal numeric value of the index, respectively. These functions are primarily for internal use, but are exported because they may be useful for users.

The replacement method also updates the tclass() and tzone() of the index to match the class and timezone of the new index, respectively. The index() method converts the internal numeric index to the class specified by the 'tclass' attribute and with the timezone specified by the 'tzone' attribute before returning the index values to the user.

The .indexXXX() functions below extract time components from the internal time index. They return values like the values of POSIXlt components.

.indexsec

0 - 61: seconds of the minute (local time)

.indexmin

0 - 59: minutes of the hour (local time)

.indexhour

0 - 23: hours of the day (local time)

.indexDate

date as seconds since the epoch (UTC not local time

.indexday

date as seconds since the epoch (UTC not local time

.indexwday

0 - 6: day of the week (Sunday - Saturday, local time)

.indexmday

1 - 31: day of the month (local time)

.indexweek

weeks since the epoch (UTC not local time

.indexmon

0 - 11: month of the year (local time)

.indexyear

years since 1900 (local time)

.indexyday

0 - 365: day of the year (local time, 365 only in leap years)

.indexisdst

1, 0, -1: Daylight Saving Time flag. Positive if Daylight Saving Time is in effect, zero if not, negative if unknown.

Changes in timezone, index class, and index format internal structure, by xts version:

Version 0.12.0:

The .indexTZ, .indexCLASS and .indexFORMAT attributes are no longer stored on xts objects, only on the index itself.

The indexTZ(), indexClass(), and indexFormat() functions (and their respective replacement methods) are deprecated in favor of their respective tzone(), tclass(), and tformat() versions. The previous versions throw a warning that they're deprecated, but they will continue to work. They will never be removed or throw an error. Ever.

The new versions are careful to look for the old attributes on the xts object, in case they're ever called on an xts object that was created prior to the attributes being added to the index itself.

You can set options(xts.warn.index.missing.tzone = TRUE) and options(xts.warn.index.missing.tclass = TRUE) to identify xts objects that do not have a 'tzone' or 'tclass' attribute on the index, even if there is a 'tzone' or 'tclass' attribute on the xts object itself. The warnings will be thrown when the object is printed. Use x <- as.xts(x) to update these objects to the new structure.

Version 0.9.8:

The index timezone is now set to "UTC" for time classes that do not have any intra-day component (e.g. days, months, quarters). Previously the timezone was blank, which meant "local time" as determined by R and the OS.

Version 0.9.2:

There are new get/set methods for the timezone, index class, and index format attributes: tzone() and, tzone<-, tclass() and tclass<-, and tformat() and tformat<-. These new functions are aliases to their indexTZ(), indexClass(), and indexFormat() counterparts.

Version 0.7.5:

The timezone, index class, and index format were added as attributes to the index itself, as 'tzone', 'tclass', and 'tformat', respectively. This is in order to remove those three attributes from the xts object, so they're only on the index itself.

The indexTZ(), indexClass(), and indexFormat() functions (and their respective replacement methods) will continue to work as in prior xts versions. The attributes on the index take priority over their respective counterparts that may be on the xts object.

Versions 0.6.4 and prior:

Objects track their timezone and index class in their '.indexTZ' and '.indexCLASS' attributes, respectively.

See Also

tformat() describes how the index values are formatted when printed, tclass() documents how xts handles the index class, and tzone() has more information about index timezone settings.

Examples

Run this code

x <- timeBasedSeq('2010-01-01/2010-01-01 12:00/H')
x <- xts(seq_along(x), x)

# the index values, converted to 'tclass' (POSIXct in this case)
index(x)
class(index(x))  # POSIXct
tclass(x)        # POSIXct

# the internal numeric index
.index(x)
# add 1 hour (3600 seconds) to the numeric index
.index(x) <- index(x) + 3600
index(x)

y <- timeBasedSeq('2010-01-01/2010-01-02 12:00')
y <- xts(seq_along(y), y)

# Select all observations in the first 6 and last 3 minutes of the
# 8th and 15th hours on each day
y[.indexhour(y) %in% c(8, 15) & .indexmin(y) %in% c(0:5, 57:59)]

i <- 0:60000
focal_date <- as.numeric(as.POSIXct("2018-02-01", tz = "UTC"))
y <- .xts(i, c(focal_date + i * 15), tz = "UTC", dimnames = list(NULL, "value"))

# Select all observations for the first minute of each hour
y[.indexmin(y) == 0]

# Select all observations on Monday
mon <- y[.indexwday(y) == 1]
head(mon)
tail(mon)
unique(weekdays(index(mon))) # check

# Disjoint time of day selections

# Select all observations between 08:30 and 08:59:59.9999  or between 12:00 and 12:14:59.99999:
y[.indexhour(y) == 8 & .indexmin(y) >= 30 | .indexhour(y) == 12 & .indexmin(x) %in% 0:14]

### Compound selections

# Select all observations for Wednesdays or Fridays between 9am and 4pm (exclusive of 4pm):
y[.indexwday(y) %in% c(3, 5) & (.indexhour(y) %in%  c(9:15))]

# Select all observations on Monday between 8:59:45 and 09:04:30:

y[.indexwday(y) == 1 & (.indexhour(y) == 8 & .indexmin(y) == 59 & .indexsec(y) >= 45 |
                        .indexhour(y) == 9 &
                        (.indexmin(y) <  4 | .indexmin(y) ==  4 & .indexsec(y) <= 30))]

i <- 0:30000
u <- .xts(i, c(focal_date + i * 1800), tz = "UTC", dimnames = list(NULL, "value"))

# Select all observations for January or February:
u[.indexmon(u) %in% c(0, 1)]

# Select all data for the 28th to 31st of each month, excluding any Fridays:
u[.indexmday(u) %in% 28:31 & .indexwday(u) != 5]

# Subset by week since origin
unique(.indexweek(u))
origin <- xts(1, as.POSIXct("1970-01-01"))
unique(.indexweek(origin))

# Select all observations in weeks 2515 to 2517.
u2 <- u[.indexweek(u) %in% 2515:2517]
head(u2); tail(u2)

# Select all observations after 12pm for day 50 and 51 in each year
u[.indexyday(u) %in% 50:51 & .indexhour(u) >= 12]

Run the code above in your browser using DataLab