Documenting
Assuming that the source_data flag is turned on, the
compiler saves all relevant documenting information collected when compiling
a source file. The provided lgtdoc tool can access this
information by using the reflection support
and generate a documentation file for each compiled entity (object, protocol,
or category) in XML format. Contents of the XML file include the entity name,
type, and compilation mode (static or dynamic), the entity relations with
other entities, and a description of any declared predicates (name,
compilation mode, scope, …). The XML documentation files can be enriched
with arbitrary user-defined information, either about an entity or about its
predicates, by using documentation directives. The lgtdoc
tool includes
POSIX and Windows scripts for converting the XML documentation files to
several final formats (such as HTML and PDF).
Logtalk supports two documentation directives for providing arbitrary user-defined information about an entity or a predicate. These two directives complement other directives that also provide important documentation information such as the mode/2 and meta_predicate/1 directives.
Entity documenting directives
Arbitrary user-defined entity information can be represented using the info/1 directive:
:- info([
Key1 is Value1,
Key2 is Value2,
...
]).
In this pattern, keys should be atoms and values should be bound terms. The following keys are predefined and may be processed specially by Logtalk tools:
comment
Comment describing the entity purpose (an atom). End the comment with a period (full stop). As a style guideline, don’t use overly long comments. If you need to provide additional details, use the
fails_if
andremarks
keys.author
Entity author(s) (an atom or a compound term
{entity}
whereentity
is the name of an XML entity in a user definedcustom.ent
file).version
Version number (a
Major:Minor:Patch
compound term) Following the Semantic Versioning guidelines is strongly advised.date
Date of last modification in ISO 8601 standard format (
Year-Month-Day
whereYear
,Month
, andDay
are integers).parameters
Parameter names and descriptions for parametric entities (a list of
Name-Description
pairs where both names and descriptions are atoms). End theDescription
with a period (full stop).parnames
Parameter names for parametric entities (a list of atoms; a simpler version of the previous key, used when parameter descriptions are deemed unnecessary).
copyright
Copyright notice for the entity source code (an atom or a compound term
{entity}
whereentity
is the name of an XML entity defined in a user definedcustom.ent
file).license
License terms for the entity source code; usually, just the license name (an atom or a compound term
{entity}
whereentity
is the name of an XML entity in a user definedcustom.ent
file). License names should whenever possible be a license identifier as specified in the SPDX standard.remarks
List of general remarks about the entity using
Topic-Text
pairs where both the topic and the text must be atoms. End theText
with a period (full stop).see_also
List of related entities (using the entity identifiers, which can be atoms or compound terms).
For example:
:- info([
version is 2:1:0,
author is 'Paulo Moura',
date is 2000-11-20,
comment is 'Building representation.',
diagram is 'UML Class Diagram #312'
]).
Use only the keywords that make sense for your application and remember that you are free to invent your own keywords. All key-value pairs can be retrieved programmatically using the reflection API and are visible to the lgtdoc tool (which includes them in the generated documentation).
Predicate documenting directives
Arbitrary user-defined predicate information can be represented using the info/2 directive:
:- info(Name/Arity, [
Key1 is Value1,
Key2 is Value2,
...
]).
The first argument can also a grammar rule non-terminal indicator,
Name//Arity
. Keys should be atoms and values should be bound terms.
The following keys are predefined and may be processed specially by
Logtalk tools:
comment
Comment describing the predicate (or non-terminal) purpose (an atom). End the comment with a period (full stop). As a style guideline, don’t use overly long comments. If you need to provide additional details, use the
remarks
key.fails_if
Comment describing failing conditions for the predicate. As a style guideline, don’t use overly long comments. If you need to provide additional details, use the
remarks
key.arguments
Names and descriptions of predicate arguments for pretty print output (a list of
Name-Description
pairs where both names and descriptions are atoms). End theDescription
with a period (full stop).argnames
Names of predicate arguments for pretty print output (a list of atoms; a simpler version of the previous key, used when argument descriptions are deemed unnecessary).
allocation
Objects where we should define the predicate. Some possible values are
container
,descendants
,instances
,classes
,subclasses
, andany
.redefinition
Describes if predicate is expected to be redefined and, if so, in what way. Some possible values are
never
,free
,specialize
,call_super_first
,call_super_last
.exceptions
List of possible exceptions throw by the predicate using
Description-Exception
pairs. The description must be an atom. The exception term must be a ground term.examples
List of typical predicate call examples using the format
Description-Goal-Bindings
. The description must be an atom with the goal sharing variables with the bindings. The variable bindings term uses the format{Variable = Term, ...}
. When there are no variable bindings, the success or failure of the predicate call should be represented by the terms{true}
or{false}
, respectively (you can also use in alternative the terms{yes}
or{no}
).remarks
List of general remarks about the predicate using
Topic-Text
pairs where both the topic and the text must be atoms. End theText
with a period (full stop).since
Version that added the predicate (
Major:Minor:Patch
).see_also
List of related predicates and non-terminals (using the predicate and non-terminal indicators).
For example:
:- info(color/1, [
comment is 'Table of defined colors.',
argnames is ['Color'],
constraint is 'Up to four visible colors allowed.',
examples is [
'Check that the color blue is defined' - color(blue) - {true}
]
]).
As with the info/1
directive, use only the keywords that make sense
for your application and remember that you are free to invent your own
keywords. All key-value pairs can also be retrieved programmatically
using the reflection API and are visible
to the lgtdoc tool (which includes them in the generated
documentation).
Describing predicates
The value of the comment
key, possibly extended with the remarks
key,
should describe a predicate purpose and, when applicable, the circumstances
under which a call may fail. Descriptions should be consistent across library
and application APIs. Some guidelines:
1. When starting the description with a verb, use the third-person singular
simple present form. For example, write 'Runs ...'
, 'Calls ...'
,
'Compares ...'
, 'Parses ...'
, 'Generates ...'
, 'Converts ...'
,
'Creates ...'
, 'Maps ...'
, 'Merges ...'
, 'Finds ...'
, etc.
2. Predicates that are pure logical relations often have descriptions starting
with 'True iff ...'
or 'True if ...'
.
3. Predicates with multiple solutions often have descriptions starting with
'Enumerates, by backtracing, all ...'
or 'Enumerates, by backtracing,
the ...'
.
4. Predicate call failure conditions often have descriptions with one or more
sentences starting with 'Fails when ...'
or 'Fails if ...'
.
If you’re not sure how best to describe a predicate, look for examples in the Logtalk libraries and developer tools APIs documentation.
Documenting predicate exceptions
As described above, the info/2
predicate directive supports an
exceptions
key that allows us to list all exceptions that may occur
when calling a predicate. For example:
:- info(check_option/1, [
comment is 'Succeeds if the option is valid. Throws an error otherwise.',
argnames is ['Option'],
exceptions is [
'``Option`` is a variable' - instantiation_error,
'``Option`` is neither a variable nor a compound term' - type_error(compound, 'Option'),
'``Option`` is a compound term but not a valid option' - domain_error(option, 'Option')
]
]).
When possible, only standard exceptions should be used. See e.g the
error handling methods section for
a full list. The argument names should be the same as those provided
in the arguments
or argnames
keys. Exceptions are usually
listed starting with instantiation and uninstantiation errors,
followed by type errors, and then domain errors. These may then be
followed by permission, existence, evaluation, representation, or
resource errors.
For each exception, use of controlled language as found e.g. in the ISO Prolog Core standard and this Handbook is advised. Some examples:
- Instantiation error when one of more arguments cannot be a variable
Argument
is a variableArgument1
andArgument2
are variables- Instantiation error when a closed list with bound elements is required
Argument
is a partial list or a list with an elementElement
which is a variable- Uninstantiation error when an argument is not a variable
Argument
is not a variable- Type error when an argument is not a variable but also not of the expected type
Argument
is neither a variable nor a TYPEArgument
is neither a partial list nor a list- Type error when an element of a list is not a variable but is not of the expected type
An element
Element
of theArgument
list is neither a variable nor a TYPE- Domain error when an argument is of the correct type but not in the expected domain
Argument
is a TYPE but not a valid DOMAINArgument
is an integer that is less than zero- Domain error when an element of a list is of the correct type but not in the expected domain
An element
Element
of theArgument
list is a TYPE but not a valid DOMAIN- Existence error when an entity of a given kind does not exist
The KIND
Argument
does not exist
Other classes of errors have a less rigid style. In case of doubt, look for examples in this Handbook, in the APIs documentation, and in standard documents.
Processing and viewing documenting files
The lgtdoc tool generates an XML documenting file per
entity. It can also generate library, directory, entity, and predicate
indexes when documenting libraries and directories. For example, assuming
the default filename extensions, a trace
object and a sort(_)
parametric object will result in trace_0.xml
and sort_1.xml
XML
files.
Each entity XML file contains references to two other files, an XML
specification file and a XSLT style-sheet file. The XML specification
file can be either a DTD file (logtalk_entity.dtd
) or an XML Scheme
file (logtalk_entity.xsd
). The XSLT style-sheet file is responsible
for converting the XML files to some desired format such as HTML or PDF.
The default names for the XML specification file and the XSL style-sheet
file are defined by the lgtdoc tool but can be
overridden by passing a list of options to the tool predicates. The
lgtdoc/xml
sub-directory in the Logtalk installation directory contains
the XML specification files described above, along with several sample XSL
style-sheet files and sample scripts for converting XML documenting files
to several formats (e.g. reStructuredText, Markdown, HTML, and PDF). For
example, assume that you want to generate the API documentation for the
types
library:
| ?- {types(loader)}.
....
| ?- {lgtdoc(loader)}.
....
| ?- lgtdoc::library(types).
...
The above queries will result in the creation of a xml_docs
in your
current directory by default. Assuming that we want to generate
Sphinx-based documentation and that we are using a POSIX operating-system,
the next steps would be:
$ cd xml_docs
$ lgt2rst -s -m
The lgt2rst
script will ask a few questions (project name, author,
version, …). After its completion, the generated HTML files will be
found in the _build/html
directory by default:
$ open _build/html/index.html
For Windows operating-systems, PowerShell (recommended) and JScript (legacy) scripts are available. For example, assuming that we want to generate HTML documentation, we could run in a PowerShell window:
cd xml_docs
lgt2html.ps1 -p saxon
When using the legacy JScript scripts, you can also use the .bat
script
alternatives:
cd xml_docs
lgt2html /p:saxon
After completion, the generated HTML files will be found in the xml_docs
directory by default.
See the NOTES
file in the tool directory for details, specially on the
XSLT processor dependencies. You may use the supplied sample files as a
starting point for generating the documentation of your Logtalk applications.
The Logtalk DTD file, logtalk_entity.dtd
, contains a reference to a
user-customizable file, custom.ent
, which declares XML entities for
source code author names, license terms, and copyright string. After
editing the custom.ent
file to reflect your personal data, you may
use the XML entities on info/1
documenting directives. For example,
assuming that the XML entities are named author, license, and
copyright we may write:
:- info([
version is 1:1:0,
author is {author},
license is {license},
copyright is {copyright}
]).
The entity references are replaced by the value of the corresponding XML entity when the XML documenting files are processed (not when they are generated; this notation is just a shortcut to take advantage of XML entities).
The lgtdoc tool supports a set of options that can be used to control the generation of the XML documentation files. See the tool documentation for details. There is also a doclet tool that allows automating the steps required to generate the documentation for an application.
Inline formatting in comments text
Inline formatting in comments text can be accomplished by using Markdown or reStructuredText syntax and converting XML documenting files to Markdown or reStructuredText files (and these, if required, to e.g. HTML, ePub, or PDF formats). Note that Markdown and reStructuredText common syntax elements are enough for most API documentation:
Mark *italic text* with one asterisk.
Mark **bold text** with two asterisks.
Mark ``monospaced text`` with two backquotes.
Rendering this block as markup gives:
Mark italic text with one asterisk. Mark bold text with two asterisks. Mark
monospaced text
with two backquotes.
As single backquotes have different purposes in Markdown (monospaced text) and reStructuredText (domain- or application-dependent meaning), never use them. This also avoids doubts if there’s an inline formatting typo in text meant to be rendered as monospaced text (usually inline code fragments).
Diagrams
The diagrams tool supports a wide range of diagrams that
can also help in documenting an application. The generated diagrams can
include URL links to both source code and API documentation. They can also
be linked, connecting for example high level diagrams to detail diagrams.
These features allow diagrams to be an effective solution for navigating and
understanding the structure and implementation of an application. This tool
uses the same reflection API as the lgtdoc
tool and thus have access to the same source data. See the tool documentation
for details.