| Ranges-class {IRanges} | R Documentation |
The Ranges virtual class is a general container for storing a set of integer ranges.
A Ranges object is a vector-like object where each element describes a "range of integer values".
A "range of integer values" is a finite set of consecutive integer values. Each range can be fully described with exactly 2 integer values which can be arbitrarily picked up among the 3 following values: its "start" i.e. its smallest (or first, or leftmost) value; its "end" i.e. its greatest (or last, or rightmost) value; and its "width" i.e. the number of integer values in the range. For example the set of integer values that are greater than or equal to -20 and less than or equal to 400 is the range that starts at -20 and has a width of 421. In other words, a range is a closed, one-dimensional interval with integer end points and on the domain of integers.
The starting point (or "start") of a range can be any integer (see
start below) but its "width" must be a non-negative integer
(see width below). The ending point (or "end") of a range is
equal to its "start" plus its "width" minus one (see end below).
An "empty" range is a range that contains no value i.e. a range that
has a null width. Depending on the context, it can be interpreted
either as just the empty set of integers or, more precisely,
as the position between its "end" and its "start" (note that
for an empty range, the "end" equals the "start" minus one).
The length of a Ranges object is the number of ranges in it, not the number of integer values in its ranges.
A Ranges object is considered empty iff all its ranges are empty.
Ranges objects have a vector-like semantic i.e. they only support single subscript subsetting (unlike, for example, standard R data frames which can be subsetted by row and by column).
The Ranges class itself is a virtual class. The following classes derive directly from the Ranges class: IRanges, IPos, NCList, PartitioningByEnd.
In the code snippets below, x, y and object are
Ranges objects. Not all the functions described below will necessarily
work with all kinds of Ranges derivatives but they should work at least
for IRanges objects.
Note that many more operations on Ranges objects are described in other
man pages of the IRanges package. See for example the man page for
intra range transformations (e.g. shift(), see
?`intra-range-methods`), or the man page for inter range
transformations (e.g. reduce(), see
?`inter-range-methods`), or the man page for
findOverlaps methods (see ?`findOverlaps-methods`),
or the man page for RangesList objects where the split
method for Ranges objects is documented.
length(x):
The number of ranges in x.
start(x):
The start values of the ranges.
This is an integer vector of the same length as x.
width(x):
The number of integer values in each range.
This is a vector of non-negative integers of the same length as x.
end(x):
start(x) + width(x) - 1L
mid(x): returns the midpoint of the range,
start(x) + floor((width(x) - 1)/2).
names(x):
NULL or a character vector of the same length as x.
update(object, ...):
Convenience method for combining multiple modifications of
object in one single call. For example
object <- update(object, start=start(object)-2L,
end=end(object)+2L)
is equivalent to
start(object) <- start(object)-2L; end(object) <- end(object)+2L.
tile(x, n, width, ...):
Splits each range in x into subranges as specified by n
(number of ranges) or width. Only one of n or width
can be specified. The return value is a IRangesList the same
length as x. Ranges with a width less than the width
argument are returned unchanged.
slidingWindows(x, width, step=1L): Generates sliding
windows within each range of x, of width width, and
starting every step positions. The return value is a
IRangesList the same length as x. Ranges with a
width less than the width argument are returned
unchanged. If the sliding windows do not exactly cover x,
the last window is partial.
isEmpty(x):
Return a logical value indicating whether x is empty or not.
as.matrix(x, ...):
Convert x into a 2-column integer matrix
containing start(x) and width(x).
Extra arguments (...) are ignored.
as.data.frame(x, row.names=NULL, optional=FALSE, ...):
Convert x into a standard R data frame object.
row.names must be NULL or a character vector giving
the row names for the data frame, and optional and any
additional argument (...) is ignored.
See ?as.data.frame for more information about these
arguments.
as.integer(x):
Convert x into an integer vector, by converting each range
into the integer sequence formed by from:to and concatenating
them together.
unlist(x, recursive = TRUE, use.names = TRUE):
Similar to as.integer(x) except can add names to elements.
x[[i]]:
Return integer vector start(x[i]):end(x[i]) denoted by i.
Subscript i can be a single integer or a character string.
x[i]:
Return a new Ranges object (of the same type as x) made of
the selected ranges.
i can be a numeric vector, a logical vector, NULL
or missing. If x is a NormalIRanges object and
i a positive numeric subscript (i.e. a numeric vector of
positive values), then i must be strictly increasing.
rep(x, times, length.out, each):
Repeats the values in x through one of the following conventions:
timesVector giving the number of times to repeat each
element if of length length(x), or to repeat the Ranges
elements if of length 1.
length.outNon-negative integer. The desired length of the output vector.
eachNon-negative integer. Each element of x is
repeated each times.
c(x, ...):
Combine x and the Ranges objects in ... together.
Any object in ... must belong to the same class as x,
or to one of its subclasses, or must be NULL.
The result is an object of the same class as x.
NOTE: Only works for IRanges (and derived) objects for now.
x * y:
The arithmetic operation x * y is for centered zooming. It
symmetrically scales the width of x by 1/y, where
y is a numeric vector that is recycled as necessary. For
example, x * 2 results in ranges with half their previous width
but with approximately the same midpoint. The ranges have been
“zoomed in”. If y is negative, it is equivalent to
x * (1/abs(y)). Thus, x * -2 would double the widths in
x. In other words, x has been “zoomed out”.
x + y: Expands the ranges in x on either side by the
corresponding value in the numeric vector y.
show(x):
By default the show method displays 5 head and 5 tail
lines. The number of lines can be altered by setting the global
options showHeadLines and showTailLines. If the
object length is less than the sum of the options, the full object
is displayed. These options affect display of IRanges,
IPos, Hits, GRanges,
GPos, GAlignments,
XStringSet objects, and more...
A Ranges object x is implicitly representing an arbitrary finite
set of integers (that are not necessarily consecutive). This set is the
set obtained by taking the union of all the values in all the ranges in
x. This representation is clearly not unique: many different
Ranges objects can be used to represent the same set of integers.
However one and only one of them is guaranteed to be normal.
By definition a Ranges object is said to be normal when its ranges are: (a) not empty (i.e. they have a non-null width); (b) not overlapping; (c) ordered from left to right; (d) not even adjacent (i.e. there must be a non empty gap between 2 consecutive ranges).
Here is a simple algorithm to determine whether x is normal:
(1) if length(x) == 0, then x is normal;
(2) if length(x) == 1, then x is normal iff
width(x) >= 1;
(3) if length(x) >= 2, then x is normal iff:
start(x)[i] <= end(x)[i] < start(x)[i+1] <= end(x)[i+1]
for every 1 <= i < length(x).
The obvious advantage of using a normal Ranges object to represent a given finite set of integers is that it is the smallest in terms of number of ranges and therefore in terms of storage space. Also the fact that we impose its ranges to be ordered from left to right makes it unique for this representation.
A special container (NormalIRanges) is provided for holding a normal IRanges object: a NormalIRanges object is just an IRanges object that is guaranteed to be normal.
Here are some methods related to the notion of normal Ranges:
isNormal(x):
Return TRUE or FALSE indicating whether x is normal or not.
whichFirstNotNormal(x):
Return NA if x is normal, or the smallest valid
indice i in x for which x[1:i] is not normal.
H. Pagès and M. Lawrence
IRanges objects (NormalIRanges objects are documented in the same man page).
The IPos class, a memory-efficient Ranges derivative for representing integer positions (i.e. integer ranges of width 1).
Ranges-comparison for comparing and ordering ranges.
findOverlaps-methods for finding/counting overlapping ranges.
intra-range-methods and inter-range-methods for intra range and inter range transformations of a Ranges derivative.
coverage-methods for computing the coverage of a set of ranges.
setops-methods for set operations on ranges.
nearest-methods for finding the nearest range neighbor.
## ---------------------------------------------------------------------
## Basic manipulation
## ---------------------------------------------------------------------
x <- IRanges(start=c(2:-1, 13:15), width=c(0:3, 2:0))
x
length(x)
start(x)
width(x)
end(x)
isEmpty(x)
as.matrix(x)
as.data.frame(x)
## Subsetting:
x[4:2] # 3 ranges
x[-1] # 6 ranges
x[FALSE] # 0 range
x0 <- x[width(x) == 0] # 2 ranges
isEmpty(x0)
## Use the replacement methods to resize the ranges:
width(x) <- width(x) * 2 + 1
x
end(x) <- start(x) # equivalent to width(x) <- 0
x
width(x) <- c(2, 0, 4)
x
start(x)[3] <- end(x)[3] - 2 # resize the 3rd range
x
## Name the elements:
names(x)
names(x) <- c("range1", "range2")
x
x[is.na(names(x))] # 5 ranges
x[!is.na(names(x))] # 2 ranges
ir <- IRanges(c(1,5), c(3,10))
ir*1 # no change
ir*c(1,2) # zoom second range by 2X
ir*-2 # zoom out 2X