With some limitations, it’s possible to use several of the Logtalk developer tools with Prolog codebases. In doing so, keep in mind that Logtalk (1) uses Prolog as a backend compiler and (2) is designed to be Prolog agnostic. The lack of sensible standardization, particularly of modules and term-expansion mechanisms, means that each Prolog system provides its own proprietary implementation of these and other concepts and mechanisms, precluding portable solutions. The Prolog adapter files used by Logtalk contain code to workaround some of these portability issues.
For best results, always use the latest Logtalk version to benefit from any recent improvements to the Logtalk compiler and Prolog backend adapter files that facilitate compiling Prolog code as Logtalk code.
This guide is most relevant for using Logtalk’s linter,
dead_code_scanner, code_metrics,
and ports_profiler tools to analyze and debug
issues in Prolog code. Other tools, such as lgtunit,
directly support Prolog code.
Note: If you’re instead considering porting a Prolog application to Logtalk,
there’s a wrapper tool that can help. In that
case, this is not the guide you’re looking for.
Plain Prolog code
Applying the Logtalk developer tools to plain Prolog code (i.e., Prolog code
not encapsulated in modules) requires wrapping the code using Logtalk objects.
A simple way of doing it is loading the
hook_objects
library and then using the
object_wrapper_hook
object. For example, assuming a foo.pl file:
| ?- {hook_objects(loader)}.
...
| ?- logtalk_compile('foo.pl', [hook(object_wrapper_hook)]).
...
In alternative, we can create a Logtalk source file per Prolog source
file, defining an object that uses an include/1 directive to load the
Prolog code. For example:
:- object(foo).
:- include('foo.pl').
:- end_object.
Keep in mind that some Prolog directives are not valid inside objects and can cause compilation errors. If that’s the case, it is best to make a copy of the Prolog files and edit them to add the object wrapper and either move the directives to outside the object or comment them.
Prolog modules
The Logtalk compiler supports compiling modules (that only make use of a common subset of directives) as objects. As, in general, a Prolog source file defines a single module, we can simply compile the Prolog files as Logtalk files. This doesn’t require any file extension changes but just passing the names of the files to the Logtalk compilation and loading built-in predicates.
Start by loading the Prolog modules as usual (modules and objects live
in different namespaces, which avoids most conflicts). This first step
allows Logtalk to query the runtime about module-exported predicates.
For example, if you use a loader.pl file to load your application:
| ?- [loader].
...
Next, try to compile the Prolog source files as Logtalk source files. For
example, assuming a directory with Prolog files using the common .pl
extension and a Prolog system providing the de facto standard forall/2
and member/2 predicates:
| ?- {os(loader)}.
...
| ?- os::directory_files('.', Files, [type(regular),extensions(['.pl']),paths(relative)]),
forall(member(File,Files), (logtalk_compile(File) -> true; true)).
...
The Logtalk compiler linter will run automatically (but note that some checks
may be turned off by default; if so, use the set_logtalk_flag/2 predicate to
turn them on). Other tools will need to be loaded and applied to the loaded
files (see the tools documentation for details). In that case, use instead:
| ?- os::directory_files('.', Files, [type(regular),extensions(['.pl']),paths(relative)]),
forall(member(File,Files), (logtalk_load(File) -> true; true)).
...
Note: As the logtalk_compile/1 and logtalk_load/1 predicates fail on
the first compilation error, we use the if-then-else control construct in the
queries above to advance to the next file in case of error. Those errors are
usually caused by Prolog proprietary extensions. How to workaround some of
those errors is discussed next.
If your Prolog application source files are found in multiple sub-directories,
you can use instead the following loader.lgt file:
:- initialization((
set_logtalk_flag(report, warnings),
logtalk_load([os(loader), types(loader)]),
logtalk_load_context(directory, Directory),
findall(
File,
( current_module(Module),
module_property(Module, file(File)),
sub_atom(File, 0, _, _, Directory)
),
Files
),
forall(
list::member(File, Files),
(logtalk_compile(File, [unknown_entities(silent)]) -> true; true)
)
)).
Change the logtalk_compile/2 to logtalk_load/2 if necessary and also play
with the available compiler options to fine-tune for your case.
Prolog code using a term-expansion mechanism
Some Prolog systems and their typical applications may make use of a
term-expansion mechanism. When the Prolog system implements this mechanism
using term_expansion/2 and goal_expansion predicates with clauses for
them collected in a selected number of modules, it’s possible to instruct
Logtalk own term-expansion mechanism to call the Prolog-defined expansions.
For example, assume that the used Prolog libraries and user modules
defining expansions store them in the user and system modules. In this
case, before attempting to compile the Prolog code as Logtalk objects, try
the following queries:
| ?- {hook_flows(loader)}.
...
| ?- set_logtalk_flag(hook, hook_set([user,system])).
...
This workaround will not work, however, when the module-defined expansion
predicates make calls to loading context predicates (such as the
prolog_load_context/2 provided by some backend Prolog compilers) that
expect the compilation context to be a module.
File loading order
Most Prolog applications load their files in a specific order. For example,
when user-defined operators are used, the file defining those operators is
loaded first. Some Prolog applications use an explicit loader file. For the
best results when trying to use the Logtalk developer tools, take into
account the loading order. If necessary, write a simple loader.lgt file
instead of using the forall/2 loop illustrated above.
Known issues
When finding a use_module/1 or use_module/2 directive in a Prolog module,
Logtalk either successfully expands it to a Logtalk-supported use_module/2
directive (which takes as its first argument a module name instead of a file
specification) or prints an error and aborts the compilation of the source
file. If you get an error, load the module file first and try again. Note
that the expansion from a Prolog use_module/1 directive to a Logtalk
use_module/2 directive will likely result in false positives when using
the dead_code_scanner tool.
The set_prolog_flag/2 directive cannot be used inside objects but is
sometimes used inside modules. The workaround is to move those directives
to outside the modules before attempting to compile the modules as objects.
Meta-predicate templates (in meta_predicate/1 directives) using : for
a meta-argument will result in a compilation error. The error results from
Logtalk not being based on a predicate-prefixing mechanism as found in
module systems and also in lack of standardization making : ambiguous, as
it can signal a goal, a closure, or an argument that will not be called as
a goal (or used to construct a goal) but still requires module-prefixing.
When the meta-argument is a goal or a closure, the workaround is to update
the meta_predicate/1 directive, replacing : by 0, in case of a goal,
or N, in case of a closure, with the integer N being the number of
additional arguments that will be added to the closure to construct a goal.
Arbitrary goals used as directives are usually flagged as errors. In most
cases, these can be wrapped using an initialization/1 directive to allow
compilation to proceed.
Clauses for term_expansion/2 and goal_expansion/2 predicates are never
used to expand clauses that follow in the same file. If that’s the case for
any module that you’re trying to compile as a Logtalk object, the only
solution is to move the expansion clauses to a separate file and compile
it before that file with the clauses to be expanded.