Breakpoints can be created using the add_breakpoint/2
built-in predicate. Its first argument should contain the
description of the breakpoint, the so called breakpoint
spec. It will return the breakpoint identifier (BID) of the
created breakpoint in its second argument. For example:
| ?- add_breakpoint(pred(foo/2), BID). % Plain spypoint for user:foo/2 added, BID=1 BID = 1
Here, we have a simple breakpoint spec, prescribing that the
debugger should stop at all ports of all invocations of the
predicate foo/2
. Thus the above goal actually creates
a plain spypoint, exactly as ?- spy foo/2.
does.
A slightly more complicated example follows:
| ?- add_breakpoint([pred(foo/2),line('/myhome/bar.pl',123)], _). % Conditional spypoint for user:foo/2 added, BID=1
This breakpoint will be activated only for those calls of
foo/2
that occur in line 123 of the Prolog program file
'/myhome/bar.pl'
. Because of the additional condition, this is
called a conditional spypoint.
The breakpoint identifier (BID) returned by
add_breakpoint/2
is an integer, assigned in increasing order,
i.e. more recent breakpoints receive higher identifier
values. When looking for applicable breakpoints, the debugger
tries the breakpoints in descending order of BIDs, i.e. the most
recent applicable breakpoint is used. Breakpoint identifiers
can be used for referring to breakpoints to be deleted, disabled
or enabled (see later).
Generally, the breakpoint spec is a pair
Tests-
Actions. Here, the Tests part describes
the conditions under which the breakpoint should be activated,
while the Actions part contains instructions on what should be
done at activation. The test part is built from tests, while the action
part from actions and tests. Test, actions and composite constructs
built from these are generally referred to as breakpoint
conditions, or simply conditions.
The action part can be omitted, and then the breakpoint spec
consists of tests only. For spypoints, the default action part is
[show(print),command(ask)]
. This instructs the debugger to print
the goal in question and then ask the user what to do next,
exactly as described in Debug Format. To illustrate other
possibilities let us explain the effect of the
[show(display),command(proceed)]
action part: this will use
display/1
for presenting the goal (just as the `d'
debugger command does, see Debug Commands), and will then proceed
with execution without stopping (i.e. the spypoint is
unleashed).