This chapter informs about command line options for GAP under UNIX
and OS X
(see Command Line Options, Advanced Features of GAP),
and features of GAP on the Macintosh (see
Running GAP under MacOS),
the .gaprc
file (see The .gaprc file),
completion files (see Completion Files),
the GAP compiler (see The Compiler, Suitability for Compilation,
Compiling Library Code),
and how to save and load a GAP workspace
(see Saving and Loading a Workspace).
When you start GAP under UNIX, you may specify a number of options on
the command-line to change the default behaviour of GAP. All these
options start with a hyphen -
, followed by a single letter. Options
must not be grouped, e.g., gap -gq
is illegal, use gap -g -q
instead. Some options require an argument, this must follow the option
and must be separated by a space, e.g., gap -m 256k
, it is not
correct to say gap -m256k
instead. Certain Boolean options (b, q, e, r,
A, D, M, N, T, X, Y) toggle the current value so that gap -b -b
is
equivalent to gap
and to gap -b -q -b -q
etc.
GAP for UNIX will distinguish between upper and lower case options.
As is described in Chapter Installing GAP (see Installing GAP),
usually you will not execute GAP directly. Instead you will
call a shell script, with the name gap
, which in turn executes GAP.
This shell script sets some options which are necessary to make GAP
work on your system. This means that the default settings mentioned below
may not be what you experience when you execute GAP on your system.
-h
-h
is mnemonic
for ``help''). GAP exits after printing the summary, all other options
are ignored.
-b
-b
toggles the state of banner display.
-q
gap>
. This is useful if you want to run GAP as
a filter with input and output redirection and want to avoid the banner
and the prompts appearing in the output file. This option may be repeated
to disable quiet mode; each -q
toggles quiet mode.
-e
ctr-D
on an empty input
line (see quit). This option should not be used when the input is a
file or pipe. This option may be repeated to toggle this behavior on and
off.
-f
-f
in this case to enable line editing. This option does not
toggle; you must use -n
to disable line editing.
-n
-n
should be used since otherwise
every input line will be echoed twice, once by Emacs and once by GAP.
This option does not toggle; you must use -f
to enable line editing.
-x
length
SizeScreen
(see SizeScreen) to alter the line length.
-y
length
SizeScreen
(see SizeScreen) to alter the number of lines.
-g
#G FULL 44580/2479kb live 57304/4392kb dead 734/4096kb free
-g -g
-g
twice, GAP prints a information message
every time a partial or full garbage collection is performed. The
message,
#G PART 9405/961kb+live 7525/1324kb+dead 2541/4096kb free
-m
memory
k
or K
it is taken as KBytes, if the last
character is m
or M
memory is taken as MBytes and if it is 'g'
or 'G' it is taken as Gigabytes.
-o
memory
k
or K
it is taken as KBytes, if the last character is
m
or M
memory is taken as MBytes and if it is 'g' or 'G' it is
taken as Gigabytes.
-K
memory
-o
above. But while the latter actually allocates more memory if
the system allows it and then prints a warning inside a break loop the -K
options tells GAP not even to try to allocate more memory. Instead GAP
just exits with an appropriate message. The default is that this feature is
switched off. You have to set it explicitly when you want to enable it.
-l
path_list
-l
is used for the first time, the only
root directory is ./
, i.e., GAP has only one root directory which is
the current directory. Usually this option is used inside a startup
script to specify where GAP is installed on the system. The -l
option can also be used by individual users to tell GAP about
privately installed modifications of the library, additional GAP
packages and so on. Section GAP Root Directory explains how several
root paths can be used to do this.
/
, but GAP will silently
add one if it is missing. If path_list does not start or end with a
semicolon, then path_list replaces the existing list of root
directories. If path_list starts with a semicolon, then path_list is
appended to the existing list of root directories. If path_list ends
with a semicolon (and does not start with one), then the new list of root
directories is the concatenation of path_list and the existing list of
root directories. After GAP has completed its startup procedure and
displays the prompt, the list of root directories can be viewed in the
variable GAPInfo.RootPaths
.
root_dir/lib/init.g
during
startup where root_dir is one of the directories in its list of root
directories. If GAP cannot find init.g
it will print the following
warning
gap: hmm, I cannot find 'lib/init.g' maybe use option '-l <gaproot>'?
-l
option.
-r
-r
tells GAP not to read the user supplied /.gaprc
files. This option may be repeated to enable reading again; each use of
-r
toggles whether to read the file.
-L
filename
-L
tells GAP to load a saved workspace. See
section Saving and Loading a Workspace.
-R
-R
tells GAP not to load a saved workspace previously
specified via the -L
option. This option does not toggle.
filename ...
The following options are in general not needed for the normal operation of GAP. They are mostly used for debugging.
-a
memory
sbrk
to get blocks of memory
from (certain) operating systems and it is required that subsequent calls
to sbrk
produce adjacent blocks of memory in this case because GAP
only wants to deal with one large block of memory. If the C function
malloc
is called for whatever reason, it is likely that sbrk
will no
longer produce adjacent blocks, therefore GAP does not use malloc
itself.
malloc
to create a
buffer when a file is opened, or for some other reason. In order to catch
these cases GAP preallocates a block of memory with malloc
which is
immediately freed. The amount preallocated can be controlled with the
-a
option. If the last character of memory is k
or K
it is taken
as KBytes and if the last character is m
or M
memory is taken as
MBytes.
-A
-B
architecture
bin
directory with in the GAP or
package root directory. The subdirectory name is determined from the
operating system, processor and compiler details when GAP (resp. the
package) is installed. Under rare circumstances, it may be necessary to
override this name, and this can be done using the -B
option.
-D
-D
option tells GAP to print short messages when it is reading
or completing files or loading modules. This option may be repeated to
toggle this behavior on and off. The message,
#I READ_GAP_ROOT: loading 'lib/kernel.g' as GAP file
lib/kernel.g
.
#I READ_GAP_ROOT: loading 'lib/kernel.g' statically
lib/kernel.g
. This compiled module was statically linked to the GAP
kernel at the time the kernel was created.
#I READ_GAP_ROOT: loading 'lib/kernel.g' dynamically
lib/kernel.g
. This compiled module was dynamically loaded to the GAP
kernel at runtime from a corresponding .so
file.
#I completing 'lib/domain.gd'
lib/domain.gd
. See
Completion Files for more information about completion of files.
-M
-N
-O
false
and fail
are identified. Use of this mode is not recommended
other than as a transitional step in porting GAP 3 code to GAP 4,
because the GAP 4 library may not work reliably in this mode. Without
the -A
option, some packages may give errors on startup. The -O
option
may be repeated to toggle this behavior on and off.
-T
quit
immediately from every break loop.
This is intended for automated testing of GAP. This option may be
repeated to toggle this behavior on and off.
-X
-Y
-i
filename
init.g
to
filename.
Additional options, -C
, -U
, -P
, -W
, -p
and -z
are used
internally in the GAP compiler and/or on specific operating systems.
This sections describes the features of GAP for MacOS that differ from those described earlier in this chapter.
Since you cannot enter command line options directly when you launch the GAP application on a Macintosh, another mechanism is being used: Hold down any of the command (apple), option, control or shift keys or space bar when launching the GAP application, e.g., by double-clicking on its icon. Please note that some keys have side effects (e.g., pressing the option key usually closes Findeer windows), and that System X behaves slightly differently from other systems.
A dialog box will open, into which you can enter the desired GAP command line options. as described in Command Line Options. For example, if you want GAP to start with a workspace of 32 megabytes, the dialog box should contain the following text:
-m 32m
Note that the dialog box may already contain settings which you have
previously saved. The OK
button accepts the command line for
the current GAP session, and the Save
button can be used to save these options
for subsequent GAP sessions. The
command line options will be saved in a text file called GAP options
in the
Preferences
folder in the system folder. You may also modify the file GAP
options
directly; note that changes only take effect the next time you launch
GAP.
There are three additional command line option on the Mac.
-z
n
-P
m
-W
m
The following command line options work differently on the Mac.
-a
-a
option has a different meaning from the one
described in Advanced Features of GAP. On the Mac, it must be used to
reserve memory for loading dynamic libraries into GAP. See The Compiler for details about dynamic libraries (and note that the PPC
version of GAP for MacOS can use dynamic libraries).
-f
, -n
-f
and -n
command line options do not have any effect on the Mac.
-e
-e
command line option enables
ctr-D
.
-o
-o
command line option should not normally be used on the Mac. The
value set by the -o
option is only used if it is lower than the size of
the workspace that would normally be available for GAP.
The file called .gaprc
on UNIX systems (see The .gaprc file) is called
gap.rc
on the Mac; it must be in the same folder as the GAP application.
All interaction between GAP and you takes place via the GAP log
window: this is where GAP prints its messages and waits for your input.
The amount of text in this window is limited (see the -W
command line option
above), so don't be surprised if old GAP messages are deleted from the
beginning of the text when this limit is reached. The reason for deleting
old lines is that otherwise GAP may run out of memory just because of
the messages it has printed.
GAP for the Mac now remembers the font and
text size (which can be set choosing Format...
in the Edit
menu)
as well as the window position of the GAP log window from one
session to the next.
Almost all of the GAP editing keys described in Section Line Editing work
on the Mac. In addition, GAP for MacOS also supports the usual
editing keys on the Mac, such as Copy and Paste, Undo, arrow keys (also
with shift
,
option
and
command
. Note that you can also move
forward and backward in the command line history by pressing
ctrl
-
arrow down and
ctrl
-
arrow up.
Note that Quit
in GAP's file menu works differently from the quit
GAP command (see quit):
Quit
in the file menu always quits the GAP application, it
cannot be used to quit from a break loop.
GAP for MacOS also contains a simple built-in text editor, which is mainly
intended to create GAP files. New
, Open...
, Save
and Close
from the
File
menu work in the usual way.
The Read...
and LogTo
commands in the File
menu work basically like the
corresponding GAP commands (see File Operations). The only difference
is that GAP will prompt you for the file with a standard Mac file opening
dialog, so you do not have to enter the path name yourself. (You will see the
file's path name in the log window afterwards). Note that if a file you want
to
read is open in GAP's built-in editor, then GAP will read the file from
the edit window, not from the disk.
If you press the shift key while choosing Read...
from the File
menu, the
menu item will change to Reread...
which will then use the GAP command Reread
(see Reread) to read the chosen file.
The Read...
command in the File
menu changes to Read
if the front window
belongs to a file in GAP's built-in editor -- choosing Read
then makes
GAP read that file -- and while the file is being read, the File
menu
item
changes to Abort Read
. You cannot close the file's window while it is being
read by GAP -- choose Abort Read
first.
Garbage collection messages, which are switched on and off by the -g
command line option (see Command Line Options) can also be switched
on and off by choosing Show garbage collections
and
Show partial collections
from the Window
menu.
If Always scroll to printout
is selected in the Window
menu, GAP will
always scroll the GAP log
window so that you can see what GAP is
currently printing. Otherwise, the GAP log window is only scrolled to
the current print position when GAP prints its prompt and
waits for you to enter a command. Note that you may see text lines disappear
even if Always scroll to printout
is off -- this happens if you are
viewing the text at the beginning of the log window and some lines are just
being deleted from the log because it has exceeded its 32000 character limit.
The contents of the Help
menu should be quite self-explanatory. Note that,
unlike in GAP 3 for the Mac, the online help is not displayed in a separate
window,
nor is the online help available while GAP is computing.
Holding down the Command (Apple) key while selecting text does the same as
selecting the text and choosing Find selection in table of contents
from
the Help
menu, holding down both Command and Option keys while selecting
tries to find the selection in the index.
When you want to refer to files or folders in GAP (for example in the
Read
, PrintTo
, AppendTo
, LogTo
commands), or have to specify files
or folders for a command line option, these files must be identified by UNIX
style path names. (Presently, GAP for MacOS also supports Mac path
names, but this may change in the future.)
Users who are familiar with UNIX path names may skip the rest of this section, noting that the working directory (i.e., folder) is the one in which the GAP application resides, and that file names on the Mac are not case sensitive.
Paths are strings used to describe where a file
is stored on a hard
disk. There are two ways for specifying UNIX path names: absolute and
relative paths. An absolute path starts with a /
, then the name of the
disk where the file is located, another /
, then a list of folders,
each containing the next one, separated by /
, and finally the name of
the file, which resides in the last folder in the list. For instance, if
your hard disk is called My HD
, and your file program.g
resides (or
should be created) in the folder programs
in the folder documents
on
My HD
, the absolute path name to that file is
/My HD/documents/programs/program.g
Relative path names work similarly, except that the starting point is
not a disk but the folder in which the GAP application program resides.
Relative path names are formed like absolute ones, except that they do
not start with a /
. Thus, if you want to access the file temp.g
in
the folder tmp
in the GAP folder, you may use the following path name:
tmp/temp.g
.
It is also possible to move upward to a parent folder: suppose that the
folder containing GAP is called applications
, which contains a folder
editor
which in turn contains the file 'program.g', then you could
access this file by the path ../editor/program.g
. The path ./
refers
to the GAP folder itself, and ../
refers to ``the folder above''.
Note also that GAP for the Mac follows (resolves) aliases to folders and files.
When you start GAP, it looks for the file with the name .gaprc
in your
home directory (on UNIX systems). On a Macintosh or a Windows system the
equivalent to the .gaprc
file is gap.rc
, and for it to be read it must
be in the same folder as the GAP application.
(Note that the file must be called gap.rc
. If you use a Windows text
editor, in particular if your default is not to show file suffixes, you
might accidentaly create a file gap.rc.txt
or gap.rc.doc
which GAP will
not recognize.)
If such a file is found it is read after
libname
/init.g
, but before any of the files mentioned on the command
line are read. You can use this file for your private customizations.
For example, if you have a file containing functions or data that you
always need, you could read this from .gaprc
. Or if you find some of
the names in the library too long, you could define abbreviations for
those names in .gaprc
. The following sample .gaprc
file does both.
Read("/usr/you/dat/mygroups.grp"); Ac := Action; AcHom := ActionHomomorphism; RepAc := RepresentativeAction;
If you have already a .gaprc
file for GAP 3, its settings might not be
compatible with GAP 4. In this case it has to be removed. On UNIX Systems
the following .gaprc
file can be used to load alternatively a .gap3rc
or
a .gap4rc
file from your home directory.
if IsBound(Permutations) then # GAP 3 Exec("echo \"READ(\\\"`pwd ~`/.gap3rc\\\");\" > /tmp/jJj"); else # GAP 4 Exec("echo \"READ(\\\"`pwd ~`/.gap4rc\\\");\" > /tmp/jJj"); fi; Read("/tmp/jJj");
The standard distribution of GAP already contains completion files so in general you do not need to create these files by yourself.
When starting, GAP reads in the whole library. As this takes some time, library files are normally condensed into completion files. These completion files contain the basic skeleton of the library but not the function bodies. When a function body is required, for example because you want to execute the corresponding function, the library file containing the function body is completed.
Completion files reduce the startup time of GAP drastically. However, this technique also means that the information stored in the completion files and the library must be consistent. If you change a library file without recreating the completion files disaster is bound to happen.
Bugfixes distributed for GAP will also update the completion files. Therefore you only need to update them if you have changed the library by yourself.
However, if you are modifying a library file a more convenient way is to
use the -X
option (see Command Line Options) that allows you (in most
cases) to use the completion files for the unchanged parts of library
files and avoids using the completion files for the changed parts. After
you have finished modifying the library files you can recreate the
completion files using:
CreateCompletionFiles( ) F
CreateCompletionFiles(
path ) F
To create completion files you must have write permissions to path
,
which defaults to the first root directory. Start GAP with the
-N
option (to suppress the reading of any existing completion files), then
execute the command CreateCompletionFiles(
path );
, where path is a
string giving a path to the home directory of GAP (the directory
containing the lib
directory).
This produces, in addition to lots of informational output, the completion files.
$ gap4 -N gap> CreateCompletionFiles(); #I converting "gap4/lib/read2.g" to "gap4/lib/read2.co" #I parsing "gap4/lib/process.gd" #I parsing "gap4/lib/listcoef.gi" ...
ARCH_IS_UNIX( ) F
tests whether GAP is running on a UNIX system.
ARCH_IS_MAC( ) F
tests whether GAP is running on a Macintosh under MacOS
ARCH_IS_WINDOWS( ) F
tests whether GAP is running on a Windows system.
The GAP compiler GAC creates C code from GAP code and then calls the system's C compiler to produce machine code from it. This can result in a speedup (see section Suitability for Compilation for more details).
To use the compiler to produce dynamically loadable modules,
call it with the -d
option:
M193 /home/ahulpke > gap4/bin/i386-ibm-linux-gcc2/gac -d test.g gap4/bin/i386-ibm-linux-gcc2/gap -C /tmp/5827_test.c test.g Init_Dynamic gcc -fpic -ansi -Wall -O2 -o /tmp/5827_test.o -I gap4/bin/i386-ibm-linux-gcc2/../../src -c /tmp/5827_test.c ld -Bshareable -x -o test.so /tmp/5827_test.o rm -f /tmp/5827_test.o rm -f /tmp/5827_test.cThis produces a file
file.so
.
LoadDynamicModule(
filename ) F
LoadDynamicModule(
filename,
crc ) F
To load a compiled file, the command LoadDynamicModule
is used. This
command loads filename as module. If given, the CRC checksum crc must
match the value of the module (see CRC Numbers).
gap> LoadDynamicModule("./test.so"); gap> CrcFile("test.g"); 2906458206 gap> LoadDynamicModule("./test.so",1); Error, <crc> mismatch (or no support for dynamic loading) called from <function>( <arguments> ) called from read-eval-loop Entering break read-eval-print loop ... you can 'quit;' to quit to outer loop, or you can 'return;' to continue brk> quit; gap> LoadDynamicModule("./test.so",2906458206);
If you want to see or modify the intermediate C code, you can also instruct
the compiler to produce only the C files by using the option -C
instead of
-d
.
There are some known problems with C code produced with the GAP compiler on 32 bit architectures and used on 64 bit architectures (and vice versa).
On some operating systems, once you have loaded a dynamic module with a certain filename, loading another with the same filename will have no effect, even if the file on disk has changed.
Typically algorithms spend large parts of their runtime only in small parts of the code. The design of GAP reflects this situation with kernel methods for many time critical calculations such as matrix or permutation arithmetic.
Compiling an algorithm whose time critical parts are already in the kernel of course will give disappointing results: Compilation will only speed up the parts that are not already in the kernel and if they make us a small part of the runtime, the overall gain is small.
Routines that benefit from compilation are those which do extensive operations with basic data types, such as lists or small integers.
The most tempting code to compile is probably the library. This section describes the mechanism used to make GAP recognize compiled versions of library files. Note however that there is no point in compiling the whole library as typically only few functions benefit from compilation as described in Section Suitability for Compilation.
All files that come with GAP are read using the internal function
READ_GAP_ROOT
. This function then checks whether a compiled version of the
file exists and if its CRC number (see CRC Numbers) matches the file.
If it does, the compiled version is loaded. Otherwise the file is read.
You can start GAP with the -D -N
option to see information printed
about this process.
To make GAP find the compiled versions, they must be
put in the bin/
systemname/compiled
directory (systemname is the name
you gave for compilation, for example i386-ibm-linux-gcc2
). They have
to be called according to the following scheme:
Suppose the file is humpty/dumpty.gi
in the GAP home
directory. Then the compiled version will be
bin/
systemname/compiled/humpty/gi/dumpty.so
. That is, the directory
hierarchy is mirrored under the compiled
directory. A further directory
level is added for the suffix of the file,
and the suffix of the compiled version of the file is set to .so
(as produced by the compiler).
For example we show how
to compile the combinat.gi
file on a Linux machine. Suppose we are in
the home directory of the gap distribution.
bin/i386-ibm-linux-gcc2/gac -d lib/combinat.gi
creates a file combinat.so
. We now put it in the right place, creating
also the necessary directories:
mkdir bin/i386-ibm-linux-gcc2/compiled mkdir bin/i386-ibm-linux-gcc2/compiled/lib mkdir bin/i386-ibm-linux-gcc2/compiled/lib/gi mv combinat.so bin/i386-ibm-linux-gcc2/compiled/lib/gi
If you now start GAP and look, for example, at the function Binomial
,
defined in combinat.gi
, you see it is indeed compiled:
gap> Print(Binomial); function ( <<arg-1>>, <<arg-2>> ) <<compiled code>> end
The command line option -M
disables the loading of compiled modules and
always reads code from the library.
CRC (cyclic redundancy check) numbers provide a certain method of doing checksums. They are used by GAP to check whether files have changed. Whenever files are ``condensed'' -- for example for completion files (see Section Completion Files) or when compiling files (see Section The compiler) -- such a checksum is computed implicitly and stored within the condensed file.
When reading a condensed version of the file instead of the original one,
the CRC checksum, which is computed via CrcFile
(see CrcFile), can be
used to check whether the original has been changed in the meantime, e.g.
gap> CrcFile("lib/morpheus.gi"); 2705743645
SaveWorkspace(
filename ) F
will save a ``snapshot'' image of the current GAP workspace in the file
filename. This image then can be loaded by another copy of GAP which
then will behave as at the point when SaveWorkspace
was called.
gap> a:=1; gap> SaveWorkspace("savefile"); true gap> quit;
SaveWorkspace
can only be used at the main gap>
prompt. It cannot
be included in the body of a loop or function, or called from a break loop.
-L
filename
A saved workspace can be loaded by starting GAP with the option -L
(see Command Line Options).
This will start GAP and load the workspace.
you@unix> gap -L savefile gap> a; 1
Please note that paths to workspaces have to be given in full, expansion of the tilde to denote a home directory will not work.
Under UNIX, it is possible to compress savefiles using gzip
. Compression
typically reduces the size of a workspace by a factor 3 or 4.
If GAP is started with a compressed savefile (omit the .gz
ending),
it will try to locate gzip
on the system and uncompress the file
automatically while reading it.
you@unix> gzip -9 savefile you@unix> gap -L savefile gap> a; 1
We cannot guarantee that saved workspaces are portable between different system architectures or over different versions of GAP or its library.
If compiled modules had been loaded into GAP before the workspace
was saved, they will be loaded into the new GAP session during the
workspace loading process. If they are not available then the load
will fail. Additional compiled modules will not be used, even if
they are available, although they may be loaded later using Reread
(see Reread).
SaveWorkspace
may sometimes produce warning messages, as in
gap> SaveWorkspace("b5"); #W bad bag id 4 found, 0 saved #W bad bag id 20 found, 0 saved true
A small number of such messages can probably be ignored (they arise
because the garbage collector may not always collect all dead objects,
and dead objects may contain data that SaveWorkspace
does not know how
to process).
GAP packages which had been loaded before the workspace was saved are loaded also when the workspace is loaded. Packages which had been available but not loaded before the workspace was saved are available also when the workspace is loaded, provided these packages have not been upgraded. Packages which have been newly installed after the workspace was saved are not available when the workspace is loaded.
GAP provides hooks for functions which are called when the prompt is to be printed and when an input line is finished.
An example of using this feature is the following function.
ColorPrompt(
bool ) F
With ColorPrompt(true);
GAP changes its user interface: The
prompts and the user input are displayed in different colors. It
also sets the variable ANSI_COLORS
to true
(which has the side
effect that some help pages are also displayed with color markup.
Switch the colored prompts off with ColorPrompt(false);
.
Note that this will only work if your terminal emulation in which
you run GAP understands the so called ANSI color escape sequences
- almost all terminal emulations on current UNIX/Linux (xterm
,
rxvt
, konsole
, ...) systems do so.
The colors shown depend on the terminal configuration and cannot be forced from an application. If your terminal follows the ANSI conventions you see the standard prompt in bold blue and the break loop prompt in bold red, as well as your input in red.
If it works for you and you like it, put the line
ColorPrompt(true);
in your .gaprc
file (see The .gaprc file).
[Top] [Up] [Previous] [Next] [Index]
GAP 4 manual
March 2006