GAP (ref) - Chapter 6: Main Loop and Break Loop

This chapter is a first of a series of chapters that describe the interactive environment in which you use GAP.

6.1 Main Loop

The normal interaction with GAP happens in the so-called read-eval-print loop. This means that you type an input, GAP first reads it, evaluates it, and then shows the result. Note that the term print may be confusing since there is a GAP function called Print (6.3-4) (see 6.3) which is in fact not used in the read-eval-print loop, but traditions are hard to break. In the following, whenever we want to express that GAP places some characters on the standard output, we will say that GAP shows something.

To signal that it is ready to accept your input, GAP shows the prompt gap>. When you see this, you know that GAP is waiting for your input.

Note that every statement must be terminated by a semicolon. You must also enter Return (i.e., strike the Return key) before GAP starts to read and evaluate your input. (The Return key may actually be marked with the word Enter and a returning arrow on your terminal.) Because GAP does not do anything until you enter Return, you can edit your input to fix typos and only when everything is correct enter Return and have GAP take a look at it (see 6.8). It is also possible to enter several statements as input on a single line. Of course each statement must be terminated by a semicolon.

It is absolutely acceptable to enter a single statement on several lines. When you have entered the beginning of a statement, but the statement is not yet complete, and you enter Return, GAP will show the partial prompt >. When you see this, you know that GAP is waiting for the rest of the statement. This happens also when you forget the semicolon ; that terminates every GAP statement. Note that when Return has been entered and the current statement is not yet complete, GAP will already evaluate those parts of the input that are complete, for example function calls that appear as arguments in another function call which needs several input lines. So it may happen that one has to wait some time for the partial prompt.

When you enter Return, GAP first checks your input to see if it is syntactically correct (see Chapter 4 for the definition of syntactically correct). If it is not, GAP prints an error message of the following form

gap> 1 * ;
Syntax error: Expression expected
1 * ;
    ^

The first line tells you what is wrong about the input, in this case the * operator takes two expressions as operands, so obviously the right one is missing. If the input came from a file (see Read (9.7-1)), this line will also contain the filename and the line number. The second line is a copy of the input. And the third line contains a caret pointing to the place in the previous line where GAP realized that something is wrong. This need not be the exact place where the error is, but it is usually quite close.

Sometimes, you will also see a partial prompt after you have entered an input that is syntactically incorrect. This is because GAP is so confused by your input, that it thinks that there is still something to follow. In this case you should enter ;Return repeatedly, ignoring further error messages, until you see the full prompt again. When you see the full prompt, you know that GAP forgave you and is now ready to accept your next –hopefully correct– input.

If your input is syntactically correct, GAP evaluates or executes it, i.e., performs the required computations (see Chapter 4 for the definition of the evaluation).

If you do not see a prompt, you know that GAP is still working on your last input. Of course, you can type ahead, i.e., already start entering new input, but it will not be accepted by GAP until GAP has completed the ongoing computation.

When GAP is ready it will usually show the result of the computation, i.e., the value computed. Note that not all statements produce a value, for example, if you enter a for loop, nothing will be printed, because the for loop does not produce a value that could be shown.

Also sometimes you do not want to see the result. For example if you have computed a value and now want to assign the result to a variable, you probably do not want to see the value again. You can terminate statements by two semicolons to suppress showing the result.

If you have entered several statements on a single line GAP will first read, evaluate, and show the first one, then read, evaluate, and show the second one, and so on. This means that the second statement will not even be checked for syntactical correctness until GAP has completed the first computation.

After the result has been shown GAP will display another prompt, and wait for your next input. And the whole process starts all over again. Note that if you have entered several statements on a single line, a new prompt will only be printed after GAP has read, evaluated, and shown the last statement.

In each statement that you enter, the result of the previous statement that produced a value is available in the variable last. The next to previous result is available in last2 and the result produced before that is available in last3.

gap> 1;2;3;
1
2
3
gap> last3 + last2 * last;
7

Also in each statement the time spent by the last statement, whether it produced a value or not, is available in the variable time (7.6-4). This is an integer that holds the number of milliseconds. Similarly the amount of memory allocated during that statement (in bytes) is stored in the variable memory_allocated (7.7-2). The variables last, last2, last3, time (7.6-4) and memory_allocated (7.7-2) are all write-protected.

6.2 Special Rules for Input Lines

The input for some GAP objects may not fit on one line, in particular big integers, long strings or long identifiers. In these cases you can still type or paste them in long single lines. For nicer display you can also specify the input on several lines. This is achieved by ending a line by a backslash or by a backslash and a carriage return character, then continue the input on the beginning of the next line. When reading this GAP will ignore such continuation backslashes, carriage return characters and newline characters. GAP also prints long strings and integers this way.

gap> n := 1234\
> 567890;
1234567890
gap> "This is a very long string that does not fit on a line \
> and is therefore continued on the next line.";
"This is a very long string that does not fit on a line and is therefo\
re continued on the next line."
gap> bla\
> bla := 5;; blabla;
5

There is a special rule about GAP prompts in input lines: In line editing mode (usual user input and GAP started without -n) in lines starting with whitespace following gap> , > or brk> this beginning part is removed. This rule is very convenient because it allows to cut and paste input from other GAP sessions or manual examples easily into your current session.

6.3 View and Print

GAP has three different operations to display or print objects: Display (6.3-6), ViewObj (6.3-5) and PrintObj (6.3-5), and these three have different purposes as follows. The first, Display (6.3-6), should print the object to the standard output in a human-readable relatively complete and verbose form. The second, ViewObj (6.3-5), should print the object to the standard output in a short and concise form, it is used in the main read-eval-print loop to display the resulting object of a computation. The third, PrintObj (6.3-5), should print the object to the standard output in a complete form which is GAP-readable if at all possible, such that reading the output into GAP produces an object which is equal to the original one.

All three operations have corresponding operations which do not print anything to standard output but return the output as a string. These are DisplayString (27.7-1), ViewString (27.7-3) and PrintString (27.7-5) (corresponding to PrintObj (6.3-5)). Additionally, there is String (27.7-6) which is very similar to PrintString (27.7-5) but does not insert control characters for line breaks.

For implementation convenience it is allowed that some of these operations have methods which delegate to some other of these operations. However, the rules for this are that a method may only delegate to another operation which appears further down in the following table:

Display (6.3-6)

ViewObj (6.3-5)

PrintObj (6.3-5)

DisplayString (27.7-1)

ViewString (27.7-3)

PrintString (27.7-5)

String (27.7-6)

Note in particular that none of the methods of the string producing operations may delegate to the corresponding printing operations. Note also that the above mentioned purposes of the different operations suggest that delegations between different operations will be sub-optimal in most scenarios.

6.3-1 Default delegations in the library

6.3-2 Recommendations for the implementation

This subsection describes what methods for printing and viewing one should implement for new GAP objects.

One should at the very least install a String (27.7-6) method to allow printing. Using the standard delegations this enables a limited form of viewing, displaying and printing.

If, for larger objects, nicer line breaks are needed, one should install a separate PrintString (27.7-5) method which puts in positions for good line breaks using the control characters \< (ASCII 1) and \> (ASCII 2).

If, for even larger objects, output performance and memory usage matters, one should install a separate PrintObj (6.3-5) method.

One should usually install a ViewString (27.7-3) method, unless the above String (27.7-6) method is good enough for ViewObj (6.3-5) purposes. Performance and memory should never matter here, so it is usually unnecessary to install a separate ViewObj (6.3-5) method.

If the type of object calls for it one should install a DisplayString (27.7-1) method. This is the case if a human readable verbose form is required.

If the performance and memory usage for Display (6.3-6) matters, one should install a separate Display (6.3-6) method.

Note that if only a String (27.7-6) method is installed, then ViewObj (6.3-5) works and ViewString (27.7-3) returns DEFAULTVIEWSTRING (27.7-4). Likewise, Display (6.3-6) works and DisplayString (27.7-1) returns DEFAULTDISPLAYSTRING (27.7-2). If you want to avoid this then install methods for these operations as well.

6.3-3 View

View shows the objects obj1, obj2... etc. in a short form on the standard output by calling the ViewObj (6.3-5) operation on each of them. View is called in the read-eval-print loop, thus the output looks exactly like the representation of the objects shown by the main loop. Note that no space or newline is printed between the objects.

6.3-4 Print

Also Print shows the objects obj1, obj2... etc. on the standard output. The difference compared to View (6.3-3) is in general that the shown form is not required to be short, and that in many cases the form shown by Print is GAP readable.

gap> z:= Z(2);
Z(2)^0
gap> v:= [ z, z, z, z, z, z, z ];
[ Z(2)^0, Z(2)^0, Z(2)^0, Z(2)^0, Z(2)^0, Z(2)^0, Z(2)^0 ]
gap> ConvertToVectorRep(v);; v;
<a GF2 vector of length 7>
gap> Print( v, "\n" );
[ Z(2)^0, Z(2)^0, Z(2)^0, Z(2)^0, Z(2)^0, Z(2)^0, Z(2)^0 ]

Another difference is that Print shows strings without the enclosing quotes, so Print can be used to produce formatted text on the standard output (see also chapter 27). Some characters preceded by a backslash, such as \n, are processed specially (see chapter 27.2). PrintTo (9.7-3) can be used to print to a file.

gap> for i in [1..5] do
>      Print( i, " ", i^2, " ", i^3, "\n" );
>    od;
1 1 1
2 4 8
3 9 27
4 16 64
5 25 125
gap> g:= SmallGroup(12,5);
<pc group of size 12 with 3 generators>
gap> Print( g, "\n" );
Group( [ f1, f2, f3 ] )
gap> View( g );  Print( "\n" );
<pc group of size 12 with 3 generators>

6.3-5 ViewObj

The functions View (6.3-3) and Print (6.3-4) actually call the operations ViewObj and PrintObj, respectively, for each argument. By installing special methods for these operations, it is possible to achieve special printing behavior for certain objects (see chapter 78). The only exceptions are strings (see Chapter 27), for which the default PrintObj and ViewObj methods as well as the function View (6.3-3) print also the enclosing doublequotes, whereas Print (6.3-4) strips the doublequotes.

The default method for ViewObj is to call PrintObj. So it is sufficient to have a PrintObj method for an object in order to View (6.3-3) it. If one wants to supply a short form for View (6.3-3), one can install additionally a method for ViewObj.

6.3-6 Display

Displays the object obj in a nice, formatted way which is easy to read (but might be difficult for machines to understand). The actual format used for this depends on the type of obj. Each method should print a newline character as last character.

gap> Display( [ [ 1, 2, 3 ], [ 4, 5, 6 ] ] * Z(5) );
 2 4 1
 3 . 2

One can assign a string to an object that Print (6.3-4) will use instead of the default used by Print (6.3-4), via SetName (12.8-1). Also, Name (12.8-2) returns the string previously assigned to the object for printing, via SetName (12.8-1). The following is an example in the context of domains.

gap> g:= Group( (1,2,3,4) );
Group([ (1,2,3,4) ])
gap> SetName( g, "C4" ); g;
C4
gap> Name( g );
"C4"

When setting up examples, in particular if for beginning users, it sometimes can be convenient to hide the structure behind a printing name. For many objects, such as groups, this can be done using SetName (12.8-1). If the objects however is represented internally, for example permutations representing group elements, this function is not applicable. Instead the function SetNameObject (6.3-7) can be used to interface with the display routines on a lower level.

6.3-7 SetNameObject

SetNameObject sets the string s as display name for object o in an interactive session. When applying View (6.3-3) to object o, for example in the system's main loop, GAP will print the string s. Calling SetNameObject for the same object o with s set to fail (20.2-1) deletes the special viewing setup. Since use of this features potentially slows down the whole print process, this function should be used sparingly.

gap> SetNameObject(3,"three");
gap> Filtered([1..10],IsPrimeInt);
[ 2, three, 5, 7 ]
gap> SetNameObject(3,fail);
gap> Filtered([1..10],IsPrimeInt);
[ 2, 3, 5, 7 ]

6.4 Break Loops

When an error has occurred or when you interrupt GAP (usually by hitting Ctrl-C) GAP enters a break loop, that is in most respects like the main read eval print loop (see 6.1). That is, you can enter statements, GAP reads them, evaluates them, and shows the result if any. However those evaluations happen within the context in which the error occurred. So you can look at the arguments and local variables of the functions that were active when the error happened and even change them. The prompt is changed from gap> to brk> to indicate that you are in a break loop.

gap> 1/0;
Rational operations: <divisor> must not be zero
not in any function
Entering break read-eval-print loop ...
you can 'quit;' to quit to outer loop, or
you can replace <divisor> via 'return <divisor>;' to continue

If errors occur within a break loop GAP enters another break loop at a deeper level. This is indicated by a number appended to brk:

brk> 1/0;
Rational operations: <divisor> must not be zero
not in any function
Entering break read-eval-print loop ...
you can 'quit;' to quit to outer loop, or
you can replace <divisor> via 'return <divisor>;' to continue
brk_02>

6.4-1 quit from a break loop

The first way to leave a break loop is to quit the break loop. To do this you enter quit; or type the eof (end of file) character, which is usually Ctrl-D except when using the -e option (see Section 3.1). Note that GAP code between quit; and the end of the input line is ignored.

In this case control returns to the break loop one level above or to the main loop, respectively. So iterated break loops must be left iteratively. Note also that if you type quit; from a gap> prompt, GAP will exit (see 6.7).

Note: If you leave a break loop with quit without completing a command it is possible (though not very likely) that data structures will be corrupted or incomplete data have been stored in objects. Therefore no guarantee can be given that calculations afterwards will return correct results! If you have been using options quitting a break loop generally leaves the options stack with options you no longer want. The function ResetOptionsStack (8.1-3) removes all options on the options stack, and this is the sole intended purpose of this function.

6.4-2 return from a break loop

The other way to leave a break loop is to return from a break loop. To do this you type return; or return obj;. If the break loop was entered because you interrupted GAP, then you can continue by typing return;. If the break loop was entered due to an error, you may have to modify the value of a variable before typing return; (see the example for IsDenseList (21.1-2)) or you may have to return an object obj (by typing: return obj;) to continue the computation; in any case, the message printed on entering the break loop will tell you which of these alternatives is possible. For example, if the break loop was entered because a variable had no assigned value, the value to be returned is often a value that this variable should have to continue the computation.

brk> return 9;  # we had tried to enter the divisor 9 but typed 0 ...
1/9
gap>

6.4-3 OnBreak

By default, when a break loop is entered, GAP prints a trace of the innermost 5 commands currently being executed. This behaviour can be configured by changing the value of the global variable OnBreak. When a break loop is entered, the value of OnBreak is checked. If it is a function, then it is called with no arguments. By default, the value of OnBreak is Where (6.4-5).

gap> OnBreak := function() Print("Hello\n"); end;
function(  ) ... end

gap> Error("!\n");
Error, !
Hello
Entering break read-eval-print loop ...
you can 'quit;' to quit to outer loop, or
you can 'return;' to continue
brk> quit;

In cases where a break loop is entered during a function that was called with options (see Chapter 8), a quit; will also cause the options stack to be reset and an Info-ed warning stating this is emitted at InfoWarning (7.4-8) level 1 (see Chapter 7.4).

Note that for break loops entered by a call to Error (6.6-1), the lines after Entering break read-eval-print loop ... and before the brk> prompt can also be customised, namely by redefining OnBreakMessage (6.4-4).

Also, note that one can achieve the effect of changing OnBreak locally. As mentioned above, the default value of OnBreak is Where (6.4-5). Thus, a call to Error (6.6-1) generally gives a trace back up to five levels of calling functions. Conceivably, we might like to have a function like Error (6.6-1) that does not trace back without globally changing OnBreak. Such a function we might call ErrorNoTraceBack and here is how we might define it. (Note ErrorNoTraceBack is not a GAP function.)

gap> ErrorNoTraceBack := function(arg) # arg is special variable that GAP
>                                      # knows to treat as list of arg's
>      local SavedOnBreak, ENTBOnBreak;
>      SavedOnBreak := OnBreak;        # save current value of OnBreak
> 
>      ENTBOnBreak := function()       # our `local' OnBreak
>      local s;
>        for s in arg do
>          Print(s);
>        od;
>        OnBreak := SavedOnBreak;      # restore OnBreak afterwards
>      end;
> 
>      OnBreak := ENTBOnBreak;
>      Error();
>    end;
function( arg... ) ... end

gap> ErrorNoTraceBack("Gidday!", " How's", " it", " going?\n");
Error, Gidday! How's it going?
Entering break read-eval-print loop ...
you can 'quit;' to quit to outer loop, or
you can 'return;' to continue
brk> quit;

gap> Error("Gidday!", " How's", " it", " going?\n");
Error, Gidday! How's it going?
Hello
Entering break read-eval-print loop ...
you can 'quit;' to quit to outer loop, or
you can 'return;' to continue
brk> quit;

Observe that the value of OnBreak before the ErrorNoTraceBack call was restored. However, we had changed OnBreak from its default value; to restore OnBreak to its default value, we should do the following.

gap> OnBreak := Where;;

6.4-4 OnBreakMessage

When a break loop is entered by a call to Error (6.6-1) the message after the Entering break read-eval-print loop ... line is produced by the function OnBreakMessage, which just like OnBreak (6.4-3) is a user-configurable global variable that is a function with no arguments.

gap> OnBreakMessage(); # By default, OnBreakMessage prints the following
you can 'quit;' to quit to outer loop, or
you can 'return;' to continue

Perhaps you are familiar with what's possible in a break loop, and so don't need to be reminded. In this case, you might wish to do the following (the first line just makes it easy to restore the default value later).

gap> NormalOnBreakMessage := OnBreakMessage;; # save the default value
gap> OnBreakMessage := function() end;        # do-nothing function
function(  ) ... end
gap> OnBreakMessage();
gap> OnBreakMessage := NormalOnBreakMessage;; # reset

With OnBreak (6.4-3) still set away from its default value, calling Error (6.6-1) as we did above, now produces:

gap> Error("!\n");
Error, !
Hello
Entering break read-eval-print loop ...
brk> quit; # to get back to outer loop

However, suppose you are writing a function which detects an error condition and OnBreakMessage needs to be changed only locally, i.e., the instructions on how to recover from the break loop need to be specific to that function. The same idea used to define ErrorNoTraceBack (see OnBreak (6.4-3)) can be adapted to achieve this. The function CosetTableFromGensAndRels (47.6-5) is an example in the GAP code where the idea is actually used.

6.4-5 Where

shows the last nr commands on the execution stack during whose execution the error occurred. If not given, nr defaults to 5. (Assume, for the following example, that after the last example OnBreak (6.4-3) has been set back to its default value.). WhereWithVars acts the same as Where while also showing the arguments and local variables of each function.

gap> StabChain(SymmetricGroup(100)); # After this we typed ^C
user interrupt at
bpt := S.orbit[1];
 called from
SiftedPermutation( S, (g * rep) ^ -1 ) called from
StabChainStrong( S.stabilizer, [ sch ], options ); called from
StabChainStrong( S.stabilizer, [ sch ], options ); called from
StabChainStrong( S, GeneratorsOfGroup( G ), options ); called from
StabChainOp( G, rec(
     ) ) called from
...
Entering break read-eval-print loop ...
you can 'quit;' to quit to outer loop, or
you can 'return;' to continue
brk> Where(2);
 called from
SiftedPermutation( S, (g * rep) ^ -1 ) called from
StabChainStrong( S.stabilizer, [ sch ], options ); called from
...

Note that the variables displayed even in the first line of the Where list (after the called from line) may be already one environment level higher and DownEnv (6.5-1) may be necessary to access them.

At the moment this backtrace does not work from within compiled code (this includes the method selection which by default is compiled into the kernel). If this creates problems for debugging, call GAP with the -M option (see 3.1) to avoid loading compiled code.

(Function calls to Info (7.4-6) and methods installed for binary operations are handled in a special way. In rare circumstances it is possible therefore that they do not show up in a Where log but the log refers to the last proper function call that happened before.)

The command line option -T to GAP disables the break loop. This is mainly intended for testing purposes and for special applications. If this option is given then errors simply cause GAP to return to the main loop.

6.5 Variable Access in a Break Loop

In a break loop access to variables of the current break level and higher levels is possible, but if the same variable name is used for different objects or if a function calls itself recursively, of course only the variable at the lowest level can be accessed.

6.5-1 DownEnv and UpEnv

DownEnv moves down nr steps in the environment and allows one to inspect variables on this level; if nr is negative it steps up in the environment again; nr defaults to 1 if not given. UpEnv acts similarly to DownEnv but in the reverse direction (the mnemonic rule to remember the difference between DownEnv and UpEnv is the order in which commands on the execution stack are displayed by Where (6.4-5)).

gap> OnBreak := function() Where(0); end;; # eliminate back-tracing on
gap>                                       # entry to break loop
gap> test:= function( n )
>    if n > 3 then Error( "!\n" ); fi; test( n+1 ); end;;
gap> test( 1 );
Error, !
Entering break read-eval-print loop ...
you can 'quit;' to quit to outer loop, or
you can 'return;' to continue
brk> Where();
 called from
test( n + 1 ); called from
test( n + 1 ); called from
test( n + 1 ); called from
<function>( <arguments> ) called from read-eval-loop
brk> n;
4
brk> DownEnv();
brk> n;
3
brk> Where();
 called from
test( n + 1 ); called from
test( n + 1 ); called from
<function>( <arguments> ) called from read-eval-loop
brk> DownEnv( 2 );
brk> n;
1
brk> Where();
 called from
<function>( <arguments> ) called from read-eval-loop
brk> DownEnv( -2 );
brk> n;
3
brk> quit;
gap> OnBreak := Where;; # restore OnBreak to its default value

Note that the change of the environment caused by DownEnv only affects variable access in the break loop. If you use return to continue a calculation GAP automatically jumps to the right environment level again.

Note also that search for variables looks first in the chain of outer functions which enclosed the definition of a currently executing function, before it looks at the chain of calling functions which led to the current invocation of the function.

gap> foo := function()
> local x; x := 1;
> return function() local y; y := x*x; Error("!!\n"); end;
> end;
function(  ) ... end
gap> bar := foo();
function(  ) ... end
gap> fun := function() local x; x := 3; bar(); end;
function(  ) ... end
gap> fun();
Error, !!
 called from
bar(  ); 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> x;
1
brk> DownEnv(1);
brk> x;
3

Here the x of foo which contained the definition of bar is found before that of fun which caused its execution. Using DownEnv we can access the x from fun.

6.6 Error and ErrorCount

6.6-1 Error

Error signals an error from within a function. First the messages messages are printed, this is done exactly as if Print (6.3-4) (see 6.3) were called with these arguments. Then a break loop (see 6.4) is entered, unless the standard error output is not connected to a terminal. You can leave this break loop with return; to continue execution with the statement following the call to Error. ErrorNoReturn (6.6-2) operates identically to Error, except it does not allow using return; to continue execution.

6.6-2 ErrorNoReturn

ErrorNoReturn signals an error from within a function. First the messages messages are printed, this is done exactly as if Print (6.3-4) (see 6.3) were called with these arguments. Then a break loop (see 6.4) is entered, unless the standard error output is not connected to a terminal. This break loop can only be exited with quit;. The function differs from Error (6.6-1) by not allowing execution to continue.

6.6-3 ErrorCount

ErrorCount returns a count of the number of errors (including user interruptions) which have occurred in the GAP session so far. The count is incremented by each error, even if GAP was started with the -T option to disable the break loop.

6.7 Leaving GAP

The normal way to terminate a GAP session is to enter either quit; (note the semicolon) or an end-of-file character (usually Ctrl-D) at the gap> prompt in the main read eval print loop.

6.7-1 QUIT

An emergency way to leave GAP is to enter QUIT at any gap> or brk> or brk_nn> prompt.

6.7-2 GapExitCode

GapExitCode sets the exit value which is returned to the operating system (or parent process) when GAP exits. This may be an integer in the range [-128..127] (other values are reduced modulo 256), or a boolean. true corresponds to the return value 0, which by convention is treated as "success". false corresponds to the return value 1, which by convention is treated as "failure". The exit value is not changed if no argument is given.

6.7-3 QuitGap

QuitGap acts similarly to the keyword QUIT, except QUIT cannot be called from a function. It exits GAP cleanly, calling any function installed using InstallAtExit (6.7-5). The optional argument ret will be passed to GapExitCode (6.7-2).

6.7-4 ForceQuitGap

ForceQuitGap is similar to QuitGap (6.7-3), except it ignores any functions installed with InstallAtExit (6.7-5), or any other functions normally run at GAP exit, such as flushing any partially outputted lines to both the screen and files, and exits GAP immediately. The optional argument ret will be passed to GapExitCode (6.7-2).

6.7-5 InstallAtExit

Before actually terminating, GAP will call (with no arguments) all of the functions that have been installed using InstallAtExit. These typically perform tasks such as cleaning up temporary files created during the session, and closing open files. If an error occurs during the execution of one of these functions, that function is simply abandoned, no break loop is entered.

gap> InstallAtExit(function() Print("bye\n"); end);
gap> quit;
bye

During execution of these functions, the global variable QUITTING will be set to true if GAP is exiting because the user typed QUIT and false otherwise. Since QUIT is considered as an emergency measure, different action may be appropriate.

6.7-6 SaveOnExitFile

If, when GAP is exiting due to a quit or end-of-file (i.e. not due to a QUIT) the variable SaveOnExitFile is bound to a string value, then the system will try to save the GAP workspace to that file, see SaveWorkspace (3.3-1).

6.8 Line Editing

In most installations GAP will be compiled to use the Gnu readline library (see the line Libs used: on GAP startup). In that case skip to the next section 6.9. (The line editing commands described in the rest of this section were available in previous versions of GAP, they will work almost the same in the standard configuration of the Gnu readline library.)

GAP allows one you to edit the current input line with a number of editing commands. Those commands are accessible either as control keys or as escape keys. You enter a control key by pressing the Ctrl key, and, while still holding the Ctrl key down, hitting another key key. You enter an escape key by hitting Esc and then hitting another key key. Below we denote control keys by Ctrl-key and escape keys by Esc-key. The case of key does not matter, i.e., Ctrl-A and Ctrl-a are equivalent.

Normally, line editing will be enabled if the input is connected to a terminal. Line editing can be enabled or disabled using the command line options -f and -n respectively (see 3.1), however this is a machine dependent feature of GAP.

Typing Ctrl-key or Esc-key for characters not mentioned below always inserts Ctrl-key resp. Esc-key at the current cursor position.

The next commands delete or kill text. The last killed text can be reinserted, possibly at a different position, with the yank command Ctrl-Y.

The Tab character, which is in fact the control key Ctrl-I, looks at the characters before the cursor, interprets them as the beginning of an identifier and tries to complete this identifier. If there is more than one possible completion, it completes to the longest common prefix of all those completions. If the characters to the left of the cursor are already the longest common prefix of all completions hitting Tab a second time will display all possible completions.

The next commands allow you to fetch previous lines, e.g., to correct typos, etc.

The four arrow keys (cursor keys) can be used instead of Ctrl-B, Ctrl-F, Ctrl-P, and Ctrl-N, respectively.

6.9 Editing using the readline library

The descriptions in this section are valid only if your GAP installation uses the readline library for command line editing. You can check by IsBound(GAPInfo.UseReadline); if this is the case.

You can use all the features of readline, as for example explained in https://tiswww.case.edu/php/chet/readline/rluserman.html. Therefore the command line editing in GAP is similar to the bash shell and many other programs. On a Unix/Linux system you may also have a manpage, try man readline.

Compared to the command line editing which was used in GAP up to version 4.4 (or compared to not using the readline library) using readline has several advantages:

6.9-1 Readline customization

You can use your readline init file (by default ~/.inputrc on Unix/Linux) to customize key bindings. If you want settings be used only within GAP you can write them between lines containing $if GAP and $endif. For a detailed documentation of the available settings and functions see here.

$if GAP
  set blink-matching-paren on
  "\C-x\C-o": dump-functions
  "\ep": kill-region
$endif

Alternatively, from within GAP the command ReadlineInitLine(line); can be used, where line is a string containing a line as in the init file.

Caveat: GAP overwrites the following keys (after reading the ~/.inputrc file): \C-g, \C-i, \C-n, \C-o, \C-p, \C-r, \C-\, \e<, \e>, Up, Down, TAB, Space, PageUp, PageDown. So, do not redefine these in your ~/.inputrc.

Note that after pressing Ctrl-v the next special character is input verbatim. This is very useful to bind keys or key sequences. For example, binding the function key F3 to the command kill-whole-line by using the sequence Ctrl-v F3 looks on many terminals like this: ReadlineInitLine("\"^[OR\":kill-whole-line");. (You can get the line back later with Ctrl-y.)

The Ctrl-g key can be used to type any unicode character by its code point. The number of the character can either be given as a count, or if the count is one the input characters before the cursor are taken (as decimal number or as hex number which starts with 0x. For example, the double stroke character ℤ can be input by any of the three key sequences Esc 8484 Ctrl-g, 8484 Ctrl-g or 0x2124 Ctrl-g.

Some terminals bind the Ctrl-s and Ctrl-q keys to stop and restart terminal output. Furthermore, sometimes Ctrl-\ quits a program. To disable this behaviour (and maybe use these keys for command line editing) you can use Exec("stty stop undef; stty start undef; stty quit undef"); in your GAP session or your gaprc file (see 3.2).

6.9-2 The command line history

GAP can save your input lines for later reuse. The keys Ctrl-p (or Up), Ctrl-n (or Down), ESC< and ESC> work as documented in 6.8, that is they scroll backward and forward in the history or go to its beginning or end. Also, Ctrl-o works as documented, it is useful for repeating a sequence of previous lines. (But Ctrl-l clears the screen as in other programs.)

The command line history can be used across several instances of GAP via the following two commands.

6.9-3 SaveCommandLineHistory

The first command saves the lines in the command line history to the file given by the string fname. The default for fname is history in the user's GAP root path GAPInfo.UserGapRoot or "~/.gap_hist" if this directory does not exist. If the optional argument app is true then the lines are appended to that file otherwise the file is overwritten.

The second command is the converse, it reads the lines from file fname. If the optional argument app is true the lines are appended to the history, else it prepends them.

By default, the command line history stores up to 1000 input lines. command line history. This number may be restricted or enlarged via via SetUserPreference("HistoryMaxLines", num); which may be set to a non negative number num to store up to num input lines or to infinity to store arbitrarily many lines. An automatic storing and restoring of the command line history can be configured via SetUserPreference("SaveAndRestoreHistory", true);.

Note that these functions are only available if your GAP is configured to use the readline library.

6.9-4 Writing your own command line editing functions

It is possible to write new command line editing functions in GAP as follows.

The functions have one argument l which is a list with five entries of the form [count, key, line, cursorpos, markpos] where count and key are the last pressed key and its count (these are not so useful here because users probably do not want to overwrite the binding of a single key), then line is a string containing the line typed so far, cursorpos is the current position of the cursor (point), and markpos the current position of the mark.

In the first three cases the result list can contain a position as a further entry, this becomes the new cursor position. Or it can contain two positions as further entries, these become the new cursor position and the new position of the mark.

Such a function can be installed as a macro for readline via InstallReadlineMacro(name, fun); where name is a string used as name of the macro and fun is a function as above. This macro can be called by a key sequence which is returned by InvocationReadlineMacro(name);.

As an example we define a function which puts double quotes around the word under or before the cursor position. The space character, the characters in "(,)", and the beginning and end of the line are considered as word boundaries. The function is then installed as a macro and bound to the key sequence Esc Q.

gap> EditAddQuotes := function(l)
>   local str, pos, i, j, new;
>   str := l[3];
>   pos := l[4];
>   i := pos;
>   while i > 1 and (not str[i-1] in ",( ") do
>     i := i-1;
>   od;
>   j := pos;
>   while IsBound(str[j]) and not str[j] in ",) " do
>     j := j+1;
>   od;
>   new := "\"";
>   Append(new, str{[i..j-1]});
>   Append(new, "\"");
>   return [i, j, new];
> end;;
gap> InstallReadlineMacro("addquotes", EditAddQuotes);
gap> invl := InvocationReadlineMacro("addquotes");;
gap> ReadlineInitLine(Concatenation("\"\\eQ\":\"",invl,"\""));;

6.10 Editing Files

In most cases, it is preferable to create longer input (in particular GAP programs) separately in an editor, and to read in the result via Read (9.7-1). Note that Read (9.7-1) by default reads from the directory in which GAP was started (respectively under Windows the directory containing the GAP binary), so you might have to give an absolute path to the file.

If you cannot create several windows, the Edit (6.10-1) command may be used to leave GAP, start an editor, and read in the edited file automatically.

6.10-1 Edit

Edit starts an editor with the file whose filename is given by the string filename, and reads the file back into GAP when you exit the editor again.

GAP will call your preferred editor if you call SetUserPreference("Editor", path); where path is the path to your editor, e.g., /usr/bin/vim. On Windows you can use edit.com.

Under macOS, you should use SetUserPreference("Editor", "open");, this will open the file in the default editor. If you call SetUserPreference("EditorOptions", ["-t"]);, the file will open in TextEdit, and SetUserPreference("EditorOptions", ["-a", "<appl>"]); will open the file using the application <appl>.

6.11 Editor Support

In the etc subdirectory of the GAP installation we provide some setup files for the editor vim.

vim is a powerful editor that understands the basic vi commands but provides much more functionality. You can find more information about it (and download it) from https://www.vim.org.

To get support for GAP syntax in vim, create in your home directory a directory .vim with subdirectories .vim/syntax and .vim/indent (If you are not using Unix, refer to the vim documentation on where to place syntax files). Then copy the file etc/vim/gap.vim to .vim/syntax/gap.vim and the file etc/vim/gap_indent.vim to .vim/indent/gap.vim.

Then edit the .vimrc file in your home directory. Add lines as in the following example:

if has("syntax")
  syntax on             " Default to no syntax highlighting
endif

" For GAP files
augroup gap
  " Remove all gap autocommands
  au!
autocmd BufRead,BufNewFile *.g,*.gi,*.gd set filetype=gap comments=s:##\ \ ,m:##\ \ ,e:##\ \ b:#

" I'm using the external program `par' for formatting comment lines starting
" with `##  '. Include these lines only when you have par installed.
  autocmd BufRead,BufNewFile *.g,*.gi,*.gd set formatprg="par w76p4s0j"
  autocmd BufWritePost,FileWritePost *.g,*.gi,*.gd set formatprg="par w76p0s0j"
augroup END

See the headers of the two mentioned files for additional comments and adjust details according to your personal taste. Send comments and suggestions to support@gap-system.org.

6.12 Changing the Screen Size

6.12-1 SizeScreen

Called with no arguments, SizeScreen returns the size of the screen as a list with two entries. The first is the length of each line, the second is the number of lines.

Called with one argument that is a list sz, SizeScreen sets the size of the screen; The first entry of sz, if bound, is the length of each line, and the second entry of sz, if bound, is the number of lines. The values for unbound entries of sz are left unaffected. The function returns the new values.

Note that those parameters can also be set with the command line options -x for the line length and -y for the number of lines (see Section 3.1).

To check/change whether line breaking occurs for files and streams see PrintFormattingStatus (10.4-8) and SetPrintFormattingStatus (10.4-8).

The line length must be between $20$ and $4096$ characters (inclusive) and the number of lines must be at least $10$. Values outside this range will be adjusted to the nearest endpoint of the range.

6.13 Teaching Mode

When using GAP in the context of (undergraduate) teaching it is often desirable to simplify some of the system output and functionality defaults (potentially at the cost of making the printing of objects more expensive). This can be achieved by turning on a teaching mode:

6.13-1 TeachingMode

When called with a boolean argument switch, this function will turn teaching mode respectively on or off.

gap> a:=Z(11)^3;
Z(11)^3
gap> TeachingMode(true);
#I  Teaching mode is turned ON
gap> a;
ZmodnZObj(8,11)
gap> TeachingMode(false);
#I  Teaching mode is turned OFF
gap> a;
Z(11)^3

6 Main Loop and Break Loop