linter

Logtalk provides a built-in linter tool that runs automatically when compiling and loading source files. The lint warnings are controlled by a set of flags. The default values for these flags are defined in the backend Prolog compiler adapter files and can be overriden from a settings file or from a source file (e.g. a loader file). These flags can be set globally using the set_logtalk_flag/2 built-in predicate. For (source file or entity) local scope, use instead the set_logtalk_flag/2 directive.

Some lint checks are turned off by default, specially when computationally expensive. Still, it’s a good idea to turn them on to check your code on a regular basis (e.g. in CI/CD pipelines).

Note that, in some cases, the linter may generate false warnings due to source code analysis limitations or special cases that, while valid when intended, usually result from programming issues. When a code rewrite is not a sensible solution to avoid the warning, the workaround is to turn off as locally as possible the flag that controls the warning.

Main linter checks

Lint checks include:

  • Missing directives (including scope, meta-predicate, dynamic, discontiguous, and multifile directives)

  • Duplicated directives, clauses, and grammar rules

  • Missing predicates (unknown messages plus calls to non-declared and non-defined predicates)

  • Calls to declared but not defined static predicates

  • Non-terminals called as predicates (instead of via the phrase/2-3 built-in methods)

  • Predicates called as non-terminals (instead of via the call//1 built-in method)

  • Non-portable predicate calls, predicate options, arithmetic function calls, directives, flags, and flag values

  • Missing arithmetic functions (with selected backends)

  • Suspicious calls (syntactically valid calls that are likely semantic errors; e.g. float comparisons using the standard arithmetic comparison operators or comparing numbers using unification)

  • Deprecated directives, predicates, arithmetic functions, control constructs, and flags

  • References to unknown entities (objects, protocols, categories, or modules)

  • Top-level shortcuts used as directives

  • Unification goals that will succeed without binding any variables

  • Unification goals that will succeed creating a cyclic term

  • Goals that are always true or always false

  • Trivial goal fails (due to no matching predicate clause)

  • Redefined built-in predicates

  • Redefined standard operators

  • Lambda expression unclassified variables and mixed up variables

  • Lambda expression with parameter variables used elsewhere in a clause

  • Singleton variables

  • If-then-else and soft cut control constructs without an else part

  • If-then-else and soft cut control constructs where the test is a unification between a variable and a ground term

  • Missing parenthesis around if-then-else and disjunction control constructs in the presence of cuts in the first argument

  • Cuts in clauses for multifile predicates

  • Missing cut in repeat loops

  • Possible non-steadfast predicate definitions

  • Non-tail recursive predicate definitions

  • Redundant calls to control constructs and built-in predicates

  • Calls to all-solutions predicates with existentially qualified variables not occurring in the qualified goal

  • Calls to all-solutions predicates with no shared variables between template and goal

  • Calls to bagof/3 and setof/3 where the goal argument contains singleton variables

  • Calls to findall/3 used to backtrack over all solutions of a goal without collecting them

  • Calls to catch/3 that catch all exceptions

  • Calls to standard predicates that have more efficient alternatives

  • Unsound calls in grammar rules

  • File, entity, predicate, and variable names not following official coding guidelines

  • Variable names that differ only on case

  • Clauses whose body is a disjunction (and that can be rewritten as multiple clauses per coding guidelines)

  • Naked meta-variables in cut-transparent control constructs

Additional lint checks are provided by the lgtunit, lgtdoc, make, and dead_code_scanner tools. For large projects, the data generated by the code_metrics tool may also be relevant in accessing code quality and suggesting code refactoring candidates.

Help on linter warnings

By loading the tutor tool, most lint warnings are expanded with explanations and suggestions on how to fix the reported issues. See also the coding guidelines for additional explanations.

Extending the linter

Experimental support for extending the linter with user-defined warnings is available using the logtalk_linter_hook/7 multifile hook predicate. For example, the format and list library objects define this hook predicate to lint calls to the format/2-3 and append/3 predicates for common errors and misuses.

Linting Prolog modules

This tool can also be applied to Prolog modules that Logtalk is able to compile as objects. For example, if the Prolog module file is named module.pl, try:

| ?- logtalk_load(module, [source_data(on)]).

Due to the lack of standardization of module systems and the abundance of proprietary extensions, this solution is not expected to work for all cases.

Linting plain Prolog files

This tool can also be applied to plain Prolog code. For example, if the Prolog file is named code.pl, simply define an object including its code:

:- object(code).
    :- include('code.pl').
:- end_object.

Save the object to an e.g. code.lgt file in the same directory as the Prolog file and then load it:

| ?- logtalk_load(code, [source_data(on)]).

In alternative, use the object_wrapper_hook provided by the hook_objects library:

| ?- logtalk_load(hook_objects(loader)).
...

| ?- logtalk_load(code, [hook(object_wrapper_hook), source_data(on)]).

With either wrapping solution, pay special attention to any compilation warnings that may signal issues that could prevent the plain Prolog from being fully checked when wrapped by an object.