Provided by: grass-doc_7.8.2-1build3_all 
NAME
r.mapcalc - Raster map calculator.
KEYWORDS
raster, algebra
SYNOPSIS
r.mapcalc
r.mapcalc --help
r.mapcalc [-sl] [expression=string] [region=string] [file=name] [seed=integer]
[--overwrite] [--help] [--verbose] [--quiet] [--ui]
Flags:
-s
Generate random seed (result is non-deterministic)
-l
List input and output maps
--overwrite
Allow output files to overwrite existing files
--help
Print usage summary
--verbose
Verbose module output
--quiet
Quiet module output
--ui
Force launching GUI dialog
Parameters:
expression=string
Expression to evaluate
region=string
The computational region that should be used.
- current uses the current region of the mapset.
- intersect computes the intersection region between
all input maps and uses the smallest resolution
- union computes the union extent of all map regions
and uses the smallest resolution
Options: current, intersect, union
Default: current
file=name
File containing expression(s) to evaluate
seed=integer
Seed for rand() function
DESCRIPTION
r.mapcalc performs arithmetic on raster map layers. New raster map layers can be created
which are arithmetic expressions involving existing raster map layers, integer or floating
point constants, and functions.
Program use
r.mapcalc expression have the form:
result = expression
where result is the name of a raster map layer to contain the result of the calculation
and expression is any legal arithmetic expression involving existing raster map layers
(except result itself), integer or floating point constants, and functions known to the
calculator. Parentheses are allowed in the expression and may be nested to any depth.
result will be created in the user’s current mapset.
As expression= is the first option, it is the default. This means that passing an
expression on the command line is possible as long as the expression is quoted and a space
is included before the first = sign. Example (’foo’ is the resulting map):
r.mapcalc "foo = 1"
or:
r.mapcalc ’foo = 1’
An unquoted expression (i.e. split over multiple arguments) won’t work, nor will omitting
the space before the = sign:
r.mapcalc ’foo=1’
Sorry, <foo> is not a valid parameter
To read command from the file, use file= explicitly, e.g.:
r.mapcalc file=file
or:
r.mapcalc file=- < file
or:
r.mapcalc file=- <<EOF
foo = 1
EOF
The formula entered to r.mapcalc by the user is recorded both in the result map title
(which appears in the category file for result) and in the history file for result.
Some characters have special meaning to the command shell. If the user is entering input
to r.mapcalc on the command line, expressions should be enclosed within single quotes. See
NOTES, below.
Computational regions in r.mapcalc
By default r.mapcalc uses the current region as computational region that was set with
g.region for processing. Sometimes it is necessary to use a region that is derived from
the raster maps in the expression to set the computational region. This is of high
importance for modules that use r.mapcalc internally to process time series of satellite
images that all have different spatial extents. A module that requires this feature is
t.rast.algebra. The region option of r.mapcalc was implemented to address this
requirement. It allows computing and using a region based on all raster maps in an
expression. Three modes are supported:
• Setting the region parameter to current will result in the use of the current
region as computational region. This is the default. The current region can be
set with g.region.
• The parameter union will force r.mapcalc to compute the disjoint union of all
regions from raster maps specified in the expression. This computed region will
then be used as computational region at runtime. The region of the mapset will
not be modified. The smallest spatial resolution of all raster maps will be used
for processing.
• The parameter intersect will force r.mapcalc to compute the intersection of all
regions from raster maps specified in the expression. This computed region will
then be used as computational region at runtime. The region of the mapset will
not be modified. The smallest spatial resolution of all raster maps will be used
for processing.
Operators and order of precedence
The following operators are supported:
Operator Meaning Type Precedence
--------------------------------------------------------------
- negation Arithmetic 12
~ one’s complement Bitwise 12
! not Logical 12
^ exponentiation Arithmetic 11
% modulus Arithmetic 10
/ division Arithmetic 10
* multiplication Arithmetic 10
+ addition Arithmetic 9
- subtraction Arithmetic 9
<< left shift Bitwise 8
>> right shift Bitwise 8
>>> right shift (unsigned) Bitwise 8
> greater than Logical 7
>= greater than or equal Logical 7
< less than Logical 7
<= less than or equal Logical 7
== equal Logical 6
!= not equal Logical 6
& bitwise and Bitwise 5
| bitwise or Bitwise 4
&& logical and Logical 3
&&& logical and[1] Logical 3
|| logical or Logical 2
||| logical or[1] Logical 2
?: conditional Logical 1
(modulus is the remainder upon division)
[1] The &&& and ||| operators handle null values differently to other operators. See the
section entitled NULL support below for more details.
The operators are applied from left to right, with those of higher precedence applied
before those with lower precedence. Division by 0 and modulus by 0 are acceptable and
give a NULL result. The logical operators give a 1 result if the comparison is true, 0
otherwise.
Raster map layer names
Anything in the expression which is not a number, operator, or function name is taken to
be a raster map layer name. Examples:
elevation
x3
3d.his
Most GRASS raster map layers meet this naming convention. However, if a raster map layer
has a name which conflicts with the above rule, it should be quoted. For example, the
expression
x = a-b
would be interpreted as: x equals a minus b, whereas
x = "a-b"
would be interpreted as: x equals the raster map layer named a-b
Also
x = 3107
would create x filled with the number 3107, while
x = "3107"
would copy the raster map layer 3107 to the raster map layer x.
Quotes are not required unless the raster map layer names look like numbers or contain
operators, OR unless the program is run non-interactively. Examples given here assume the
program is run interactively. See NOTES, below.
r.mapcalc will look for the raster map layers according to the user’s current mapset
search path. It is possible to override the search path and specify the mapset from which
to select the raster map layer. This is done by specifying the raster map layer name in
the form:
name@mapset
For example, the following is a legal expression:
result = x@PERMANENT / y@SOILS
The mapset specified does not have to be in the mapset search path. (This method of
overriding the mapset search path is common to all GRASS commands, not just r.mapcalc.)
The neighborhood modifier
Maps and images are data base files stored in raster format, i.e., two-dimensional
matrices of integer values. In r.mapcalc, maps may be followed by a neighborhood modifier
that specifies a relative offset from the current cell being evaluated. The format is
map[r,c], where r is the row offset and c is the column offset. For example, map[1,2]
refers to the cell one row below and two columns to the right of the current cell,
map[-2,-1] refers to the cell two rows above and one column to the left of the current
cell, and map[0,1] refers to the cell one column to the right of the current cell. This
syntax permits the development of neighborhood-type filters within a single map or across
multiple maps.
Raster map layer values from the category file
Sometimes it is desirable to use a value associated with a category’s label instead of the
category value itself. If a raster map layer name is preceded by the @ operator, then the
labels in the category file for the raster map layer are used in the expression instead of
the category value.
For example, suppose that the raster map layer soil.ph (representing soil pH values) has a
category file with labels as follows:
cat label
------------------
0 no data
1 1.4
2 2.4
3 3.5
4 5.8
5 7.2
6 8.8
7 9.4
Then the expression:
result = @soils.ph
would produce a result with category values 0, 1.4, 2.4, 3.5, 5.8, 7.2, 8.8 and 9.4.
Note that this operator may only be applied to raster map layers and produces a floating
point value in the expression. Therefore, the category label must start with a valid
number. If the category label is integer, it will be represented by a floating point
number. I the category label does not start with a number or is missing, it will be
represented by NULL (no data) in the resulting raster map.
Grey scale equivalents and color separates
It is often helpful to manipulate the colors assigned to map categories. This is
particularly useful when the spectral properties of cells have meaning (as with imagery
data), or when the map category values represent real quantities (as when category values
reflect true elevation values). Map color manipulation can also aid visual recognition,
and map printing.
The # operator can be used to either convert map category values to their grey scale
equivalents or to extract the red, green, or blue components of a raster map layer into
separate raster map layers.
result = #map
converts each category value in map to a value in the range 0-255 which represents the
grey scale level implied by the color for the category. If the map has a grey scale color
table, then the grey level is what #map evaluates to. Otherwise, it is computed as:
0.10 * red + 0.81 * green + 0.01 * blue
Alternatively, you can use:
result = y#map
to use the NTSC weightings:
0.30 * red + 0.59 * green + 0.11 * blue
Or, you can use:
result = i#map
to use equal weightings:
0.33 * red + 0.33 * green + 0.33 * blue
The # operator has three other forms: r#map, g#map, b#map. These extract the red, green,
or blue components in the named raster map, respectively. The GRASS shell script r.blend
extracts each of these components from two raster map layers, and combines them by a
user-specified percentage. These forms allow color separates to be made. For example, to
extract the red component from map and store it in the new 0-255 map layer red, the user
could type:
red = r#map
To assign this map grey colors type:
r.colors map=red color=rules
black
white
To assign this map red colors type:
r.colors map=red color=rules
black
red
Functions
The functions currently supported are listed in the table below. The type of the result
is indicated in the last column. F means that the functions always results in a floating
point value, I means that the function gives an integer result, and * indicates that the
result is float if any of the arguments to the function are floating point values and
integer if all arguments are integer.
function description type
---------------------------------------------------------------------------
abs(x) return absolute value of x *
acos(x) inverse cosine of x (result is in degrees) F
asin(x) inverse sine of x (result is in degrees) F
atan(x) inverse tangent of x (result is in degrees) F
atan(x,y) inverse tangent of y/x (result is in degrees) F
ceil(x) the smallest integral value not less than x *
cos(x) cosine of x (x is in degrees) F
double(x) convert x to double-precision floating point F
eval([x,y,...,]z) evaluate values of listed expr, pass results to z
exp(x) exponential function of x F
exp(x,y) x to the power y F
float(x) convert x to single-precision floating point F
floor(x) the largest integral value not greater than x *
graph(x,x1,y1[x2,y2..]) convert the x to a y based on points in a graph F
graph2(x,x1[,x2,..],y1[,y2..])
alternative form of graph() F
if decision options: *
if(x) 1 if x not zero, 0 otherwise
if(x,a) a if x not zero, 0 otherwise
if(x,a,b) a if x not zero, b otherwise
if(x,a,b,c) a if x > 0, b if x is zero, c if x < 0
int(x) convert x to integer [ truncates ] I
isnull(x) check if x = NULL
log(x) natural log of x F
log(x,b) log of x base b F
max(x,y[,z...]) largest value of those listed *
median(x,y[,z...]) median value of those listed *
min(x,y[,z...]) smallest value of those listed *
mode(x,y[,z...]) mode value of those listed *
nmax(x,y[,z...]) largest value of those listed, excluding NULLs *
nmedian(x,y[,z...]) median value of those listed, excluding NULLs *
nmin(x,y[,z...]) smallest value of those listed, excluding NULLs *
nmode(x,y[,z...]) mode value of those listed, excluding NULLs *
not(x) 1 if x is zero, 0 otherwise
pow(x,y) x to the power y *
rand(a,b) random value x : a <= x < b *
round(x) round x to nearest integer I
round(x,y) round x to nearest multiple of y
round(x,y,z) round x to nearest y*i+z for some integer i
sin(x) sine of x (x is in degrees) F
sqrt(x) square root of x F
tan(x) tangent of x (x is in degrees) F
xor(x,y) exclusive-or (XOR) of x and y I
Internal variables:
row() current row of moving window I
col() current col of moving window I
nrows() number of rows in computation region I
ncols() number of columns in computation region I
x() current x-coordinate of moving window F
y() current y-coordinate of moving window F
ewres() current east-west resolution F
nsres() current north-south resolution F
area() area of current cell in square meters F
null() NULL value
Note, that the row() and col() indexing starts with 1.
Floating point values in the expression
Floating point numbers are allowed in the expression. A floating point number is a number
which contains a decimal point:
2.3 12.0 12. .81
Floating point values in the expression are handled in a special way. With arithmetic and
logical operators, if either operand is float, the other is converted to float and the
result of the operation is float. This means, in particular that division of integers
results in a (truncated) integer, while division of floats results in an accurate floating
point value. With functions of type * (see table above), the result is float if any
argument is float, integer otherwise.
Note: If you calculate with integer numbers, the resulting map will be integer. If you
want to get a float result, add the decimal point to integer number(s).
If you want floating point division, at least one of the arguments has to be a floating
point value. Multiplying one of them by 1.0 will produce a floating-point result, as will
using float():
r.mapcalc "ndvi = float(lsat.4 - lsat.3) / (lsat.4 + lsat.3)"
NULL support
• Division by zero should result in NULL.
• Modulus by zero should result in NULL.
• NULL-values in any arithmetic or logical operation should result in NULL.
(however, &&& and ||| are treated specially, as described below).
• The &&& and ||| operators observe the following axioms even when x is NULL:
x &&& false == false
false &&& x == false
x ||| true == true
true ||| x == true
• NULL-values in function arguments should result in NULL (however, if(), eval() and
isnull() are treated specially, as described below).
• The eval() function always returns its last argument
• The situation for if() is:
if(x)
NULL if x is NULL; 0 if x is zero; 1 otherwise
if(x,a)
NULL if x is NULL; a if x is non-zero; 0 otherwise
if(x,a,b)
NULL if x is NULL; a if x is non-zero; b otherwise
if(x,n,z,p)
NULL if x is NULL; n if x is negative;
z if x is zero; p if x is positive
• The (new) function isnull(x) returns: 1 if x is NULL; 0 otherwise. The (new)
function null() (which has no arguments) returns an integer NULL.
• Non-NULL, but invalid, arguments to functions should result in NULL.
Examples:
log(-2)
sqrt(-2)
pow(a,b) where a is negative and b is not an integer
NULL support: Please note that any math performed with NULL cells always results in a NULL
value for these cells. If you want to replace a NULL cell on-the-fly, use the isnull()
test function in a if-statement.
Example: The users wants the NULL-valued cells to be treated like zeros. To add maps A and
B (where B contains NULLs) to get a map C the user can use a construction like:
C = A + if(isnull(B),0,B)
NULL and conditions:
For the one argument form:
if(x) = NULL if x is NULL
if(x) = 0 if x = 0
if(x) = 1 otherwise (i.e. x is neither NULL nor 0).
For the two argument form:
if(x,a) = NULL if x is NULL
if(x,a) = 0 if x = 0
if(x,a) = a otherwise (i.e. x is neither NULL nor 0).
For the three argument form:
if(x,a,b) = NULL if x is NULL
if(x,a,b) = b if x = 0
if(x,a,b) = a otherwise (i.e. x is neither NULL nor 0).
For the four argument form:
if(x,a,b,c) = NULL if x is NULL
if(x,a,b,c) = a if x > 0
if(x,a,b,c) = b if x = 0
if(x,a,b,c) = c if x < 0
More generally, all operators and most functions return NULL if *any* of their arguments
are NULL.
The functions if(), isnull() and eval() are exceptions.
The function isnull() returns 1 if its argument is NULL and 0 otherwise. If the user
wants the opposite, the ! operator, e.g. "!isnull(x)" must be used.
All forms of if() return NULL if the first argument is NULL. The 2, 3 and 4 argument forms
of if() return NULL if the "selected" argument is NULL, e.g.:
if(0,a,b) = b regardless of whether a is NULL
if(1,a,b) = a regardless of whether b is NULL
eval() always returns its last argument, so it only returns NULL if the last argument is
NULL.
Note: The user cannot test for NULL using the == operator, as that returns NULL if either
or both arguments are NULL, i.e. if x and y are both NULL, then "x == y" and "x != y" are
both NULL rather than 1 and 0 respectively.
The behaviour makes sense if the user considers NULL as representing an unknown quantity.
E.g. if x and y are both unknown, then the values of "x == y" and "x != y" are also
unknown; if they both have unknown values, the user doesn’t know whether or not they both
have the same value.
NOTES
Usage from command line
Extra care must be taken if the expression is given on the command line. Some characters
have special meaning to the UNIX shell. These include, among others:
* ( ) > & |
It is advisable to put single quotes around the expression; e.g.:
’result = elevation * 2’
Without the quotes, the *, which has special meaning to the UNIX shell, would be altered
and r.mapcalc would see something other than the *.
Multiple computations
In general, it’s preferable to do as much as possible in each r.mapcalc command. E.g.
rather than:
r.mapcalc "$GIS_OPT_OUTPUT.r = r#$GIS_OPT_FIRST * .$GIS_OPT_PERCENT + (1.0 - .$GIS_OPT_PERCENT) * r#$GIS_OPT_SECOND"
r.mapcalc "$GIS_OPT_OUTPUT.g = g#$GIS_OPT_FIRST * .$GIS_OPT_PERCENT + (1.0 - .$GIS_OPT_PERCENT) * g#$GIS_OPT_SECOND"
r.mapcalc "$GIS_OPT_OUTPUT.b = b#$GIS_OPT_FIRST * .$GIS_OPT_PERCENT + (1.0 - .$GIS_OPT_PERCENT) * b#$GIS_OPT_SECOND"
use:
r.mapcalc <<EOF
$GIS_OPT_OUTPUT.r = r#$GIS_OPT_FIRST * .$GIS_OPT_PERCENT + (1.0 - .$GIS_OPT_PERCENT) * r#$GIS_OPT_SECOND
$GIS_OPT_OUTPUT.g = g#$GIS_OPT_FIRST * .$GIS_OPT_PERCENT + (1.0 - .$GIS_OPT_PERCENT) * g#$GIS_OPT_SECOND
$GIS_OPT_OUTPUT.b = b#$GIS_OPT_FIRST * .$GIS_OPT_PERCENT + (1.0 - .$GIS_OPT_PERCENT) * b#$GIS_OPT_SECOND
EOF
as the latter will read each input map only once.
Backwards compatibility
For the backwards compatibility with GRASS 6, if no options are given, it manufactures
file=- (which reads from stdin), so you can continue to use e.g.:
r.mapcalc < file
or:
r.mapcalc <<EOF
foo = 1
EOF
But unless you need compatibility with previous GRASS GIS versions, use file= explicitly,
as stated above.
When the map name contains uppercase letter(s) or a dot which are not allowed to be in
module option names, the r.mapcalc command will be valid also without quotes:
r.mapcalc elevation_A=1
r.mapcalc elevation.1=1
However, this syntax is not recommended as quotes as stated above more safe. Using quotes
is both backwards compatible and valid in future.
Interactive input in command line
For formulas that the user enters from standard input (rather than from the command line),
a line continuation feature now exists. If the user adds a backslash to the end of an
input line, r.mapcalc assumes that the formula being entered by the user continues on to
the next input line. There is no limit to the possible number of input lines or to the
length of a formula.
If the r.mapcalc formula entered by the user is very long, the map title will contain only
some of it, but most (if not all) of the formula will be placed into the history file for
the result map.
Raster MASK handling
r.mapcalc follows the common GRASS behavior of raster MASK handling, so the MASK is only
applied when reading an existing GRASS raster map. This implies that, for example, the
command:
r.mapcalc "elevation_exaggerated = elevation * 3"
create a map respecting the masked pixels if MASK is active.
However, when creating a map which is not based on any map, e.g. a map from a constant:
r.mapcalc "base_height = 200.0"
the created raster map is limited only by a computation region but it is not affected by
an active MASK. This is expected because, as mentioned above, MASK is only applied when
reading, not when writing a raster map.
If also in this case the MASK should be applied, an if() statement including the MASK
should be used, e.g.:
r.mapcalc "base_height = if(MASK, 200.0, null())"
When testing MASK related expressions keep in mind that when MASK is active you don’t see
data in masked areas even if they are not NULL. See r.mask for details.
eval function
If the output of the computation should be only one map but the expression is so complex
that it is better to split it to several expressions, the eval function can be used:
r.mapcalc << EOF
eval(elev_200 = elevation - 200, \
elev_5 = 5 * elevation, \
elev_p = pow(elev_5, 2))
elevation_result = (0.5 * elev_200) + 0.8 * elev_p
EOF
This example uses unix-like << EOF syntax to provide input to r.mapcalc.
Note that the temporary variables (maps) are not created and thus it does not matter
whether they exists or not. In the example above, if map elev_200 exists it will not be
overwritten and no error will be generated. The reason is that the name elev_200 now
denotes the temporary variable (map) and not the existing map. The following parts of the
expression will use the temporary elev_200 and the existing elev_200 will be left intact
and will not be used. If a user want to use the existing map, the name of the temporary
variable (map) must be changed.
Using the same map for input and output results
A map cannot be used both as an input and as an output as in this invalid expression
oldmap = oldmap + 1, instead a subsequent rename using g.rename is needed when the same
name is desired:
r.mapcalc "newmap = oldmap + 1"
g.rename raster=newmap,oldmap
Random number generator initialization
The pseudo-random number generator used by the rand() function can be initialised to a
specific value using the seed option. This can be used to replicate a previous
calculation.
Alternatively, it can be initialised from the system time and the PID using the -r flag.
This should result in a different seed being used each time.
In either case, the seed will be written to the map’s history, and can be seen using
r.info.
If you want other people to be able to verify your results, it’s preferable to use the
seed option to supply a seed which is either specified in the script or generated from a
determenistic process such as a pseudo-random number generator given an explicit seed.
Note that the rand() function will generate a fatal error if neither the seed option nor
the -s flag are given.
EXAMPLES
To compute the average of two raster map layers a and b:
ave = (a + b)/2
To form a weighted average:
ave = (5*a + 3*b)/8.0
To produce a binary representation of the raster map layer a so that category 0 remains 0
and all other categories become 1:
mapmask = a != 0
This could also be accomplished by:
mapmask = if(a)
To mask raster map layer b by raster map layer a:
result = if(a,b)
To change all values below 5 to NULL:
newmap = if(map<5, null(), 5)
To create a map with random values in a defined range (needs either the usage of -s flag
or the seed parameter). The precision of the input values determines the output precision
(the resulting raster map type):
# write result as integer map (CELL)
random_int = rand(-100,100)
# write result as double precision floating point map (DCELL)
random_dcell = rand(-100.0,100.0)
# write result as single precision floating point map (FCELL)
random_fcell = float(rand(-100.0,100.0))
The graph() function allows users to specify a x-y conversion using pairs of x,y
coordinates. In some situations a transformation from one value to another is not easily
established mathematically, but can be represented by a 2-D graph and then linearly
interpolated. The graph() function provides the opportunity to accomplish this. An x-axis
value is provided to the graph function along with the associated graph represented by a
series of x,y pairs. The x values must be monotonically increasing (each larger than or
equal to the previous). The graph function linearly interpolates between pairs. Any x
value lower the lowest x value (i.e. first) will have the associated y value returned.
Any x value higher than the last will similarly have the associated y value returned.
Consider the request:
newmap = graph(map, 1,10, 2,25, 3,50)
X (map) values supplied and y (newmap) values returned:
0, 10
1, 10
1.5, 17.5
2.9, 47.5
4, 50
100, 50
KNOWN ISSUES
The result variable on the left hand side of the equation should not appear in the
expression on the right hand side.
mymap = if( mymap > 0, mymap, 0)
Any maps generated by a r.mapcalc command only exist after the entire command has
completed. All maps are generated concurrently, row-by-row (i.e. there is an implicit "for
row in rows {...}" around the entire expression). Thus the #, @, and [ ] operators cannot
be used on a map generated within same r.mapcalc command run. Consequently, the following
(strikethrough code) does not work:
newmap = oldmap * 3.14
othermap = newmap[-1, 0] / newmap[1, 0]
Continuation lines must end with a \ and have no trailing white space (blanks or tabs). If
the user does leave white space at the end of continuation lines, the error messages
produced by r.mapcalc will be meaningless and the equation will not work as the user
intended. This is particularly important for the eval() function.
Currently, there is no comment mechanism in r.mapcalc. Perhaps adding a capability that
would cause the entire line to be ignored when the user inserted a # at the start of a
line as if it were not present, would do the trick.
The function should require the user to type "end" or "exit" instead of simply a blank
line. This would make separation of multiple scripts separable by white space.
r.mapcalc does not print a warning in case of operations on NULL cells. It is left to the
user to utilize the isnull() function.
SEE ALSO
g.region, r.bitpattern, r.blend, r.colors, r.fillnulls, r.mapcalc.simple
REFERENCES
r.mapcalc: An Algebra for GIS and Image Processing, by Michael Shapiro and Jim Westervelt,
U.S. Army Construction Engineering Research Laboratory (March/1991).
Performing Map Calculations on GRASS Data: r.mapcalc Program Tutorial, by Marji Larson,
Michael Shapiro and Scott Tweddale, U.S. Army Construction Engineering Research Laboratory
(December 1991)
Grey scale conversion is based on the C.I.E. x,y,z system where y represents luminance.
See "Fundamentals of Digital Image Processing," by Anil K. Jain (Prentice Hall, NJ, 1989;
p 67).
AUTHORS
Michael Shapiro, U.S.Army Construction Engineering Research Laboratory
Glynn Clements
SOURCE CODE
Available at: r.mapcalc source code (history)
Main index | Raster index | Topics index | Keywords index | Graphical index | Full index
© 2003-2019 GRASS Development Team, GRASS GIS 7.8.2 Reference Manual