Coding conventions for SIS developers

This page describes some coding conventions applied in Apache SIS development. Note that the recommended code patterns for SIS users apply also to SIS development.

License header

All Java source files (*.java) shall begin with the current ASF license header as described in ASF Source Header. Properties source files (*.properties) used as inputs to some processor (e.g. the resource compiler) shall have the same license header, but with lines prefixed by # instead of *. Properties files distributed as-is in the JAR files can summarize the license on a single line for saving space, as below:

# Licensed to the Apache Software Foundation (ASF) under one or more contributor license agreements.

Public API

The public API is made of all public types in all exported packages, i.e. packages declared in non-qualified exports statements in the module-info.java file. All other classes are for SIS usage only and may change without warning in any future release. Those classes are excluded from Javadoc and normally not accessible to users. Contrarily to previous SIS versions, there is no longer any particular naming convention for internal package. They may or may not have internal in their name.

Substitution for non-available JDK classes

When using a JDK 12+ class that does not exist on JDK 11, define a class such as JDK12 in the org.apache.sis.pending.jdk package with the minimal amount of needed functionalities, provided that it can be done with reasonable effort. This packages is internal (non-exported).

Naming convention

Classes implementing GeoAPI interfaces usually (but not always) begin with Abstract, Default, Simple or General prefix.

• The Abstract prefix is used when a class is abstract according ISO specifications — it may or may not be be abstract in the Java sense.
• The General prefix is used when an implementation is designed for use in the general case, as opposed to other implementations specialized for a fixed number of dimensions or other conditions.
• Implementations specialized for a fixed number of dimensions are suffixed with 1D, 2D, 3D or 4D rather than being prefixed.

Example: the GeneralEnvelope class is an implementation of Envelope interface for the multi-dimensional case. The Envelope2D class is another implementation of the same interface specialized for the two-dimensional case.

Member fields do not have any particular prefix (no m_ prefix).

Code formatting

Apache SIS uses the standard Java conventions, except the 80 characters line length restriction which is relaxed. The conventions listed below are guidelines. Some exceptions to coding conventions can occur but should be rare. For making merges between branches easier, refrain from doing massive code reformatting unless:

• the modified files do not yet exist on the other branches;
• or the modified lines are known to be identical on all active branches (merges work well in such cases);
• or the committer is willing to resolve the merge conflicts.

Import statements

There is currently no strict rule about the order of import statements in Apache SIS code base, except one: if a class is different in the main, geoapi-3.1 and geoapi-4.0 branches, and if those differences imply different import statements, then the imports that are different should be grouped last with a comment. Example:

import java.io.File;
import java.util.List;
import org.opengis.metadata.Metadata;

// Specific to the geoapi-3.1 and geoapi-4.0 branches:
import org.opengis.filter.Filter;
import org.opengis.filter.Expression;

The purpose is to reduce the number of conflicts during the merges between branches. The import statements can be rearranged automatically by the ReorganizeImports class in buildSrc. This tool requires checkouts of all three branches (main, geoapi-3.1 and geoapi-4.0) in order to identify which imports are branch-specific.

Spaces, brackets and keywords

• Indentation:
  • Use 4 space indents (not tab) for Java files.
  • Use 2 space indents for XML files, because ISO/OGC XML schemas tend to have a very deep structure.
• Spaces:
  • Put a space after if, else, for, try, catch and similar keywords (not after method names).
  • Remove all trailing whitespaces at the end of lines.
    • Eclipse users can use the Source - Cleanup option to accomplish this.
    • NetBeans users can use the use the Source - Remove trailing spaces on a file-by-file basis, or set the Preferences - Editor - On Save - Remove trailing whitespaces option.
• Line wrapping:
  • Use 120-column line width for Java code and Javadoc.
  • Some exceptions to above rule may exist for preserving tabular structures, but should be rare.
• Curly brackets:
  • The { and } brackets are mandatory for if, else, while and other blocks, except if the instruction after the keyword is on the same line (e.g. else if).
• Member declarations:
  • Keywords should appear in the “customary order” of the Java Language Specification: public, protected or private, then abstract or static, finally final.
  • The var keyword should be used only if the type is obvious on the same line, e.g. before a new statement or a cast.

Exceptions to coding conventions

Many mathematical operations in a two-dimensional space (or more) have symmetry. Formatting the code in a way that emphase their symmetrical nature, for example aligning identical terms in columns, can help to understand the overall pattern and to identify bugs. Example:

if (x < xmin) xmin = x;
if (x > xmax) xmax = x;
if (y < ymin) ymin = y;
if (y > ymax) ymax = y;

The var keyword may also be used aven when the type is not obvious if doing so simplify merges between branches. For example, a method returning Feature on the GeoAPI 3.1 branch may return AbstractFeature on the GeoAPI 3.0 branch. The decision to use standard or tabular format is made on a case-by-case basis. Of course, tabular format shall not be abused.

Documentation formatting

Javadoc comments are written in HTML, not Markdown, both for historical reasons and because HTML allows richer semantic. Apache SIS uses the standard Javadoc conventions, except for the 80 characters line length restriction which is relaxed. Javadoc lines should not exceed 120 characters, but exceptions to this rule may exist for preserving the structure of <table> elements.

Javadoc annotations

SIS uses standard javadoc annotations. The meaning of some tags are refined as below:

• @since - the SIS version when the annotated element (class, method, etc.) was first introduced in public API.
• @version - the last SIS version when the code of the annotated class got a significant change.
• @author - developer name in FirstName LastName (Organization) format. A separated @author tag is added for each developer. The intent is to allow other developers to know to who to ask questions if needed.

The @since and @version Javadoc tags should be used only on classes, interfaces, enumerations, methods or fields that are part of public API. They should not be put on package-private classes or classes in non-exported packages. The reason is that non-public classes can be moved, splitted or merged without warning, which gives confusing meaning to the @since tag. Furthermore, restricting the use of those tags to public API is also a way to remind developers that the class that they are editing is part of public API, so backward-compatibility concerns apply.

In addition, the Java code in buildSrc provides the following custom javadoc taglets:

HTML elements

HTML tags and entities shall be used only when there is no equivalent Javadoc tag. For example:

• Instead of “<code>Foo</code>”, use “{@code Foo}”.
• Instead of “a < b < c”, use “{@literal a < b < c}”.
• Instead of “<pre>Foo</pre>” for a Java listing, use “{@snippet lang=java : Foo}”.

When many HTML tags produce the same visual effect, choose the one with proper semantic. For example, for formatting a text in italic, choose the most appropriate of the following tags:

• <em> for emphasis. A screen reader may pronounce the words using verbal stress.
• <var> for a variable to show like a mathematical symbol.
• <dfn> for introducing a word defined by the nearby sentences.
• <cite> for the title of a document, in particular an OGC/ISO standard. Apache SIS uses also this tag for the name of a geodetic object in the EPSG geodetic database, in which case the object definition is considered as a document. This tag can also be used for section titles.
• <i> for rendering in italic for any reason other than the above reasons.

Paragraphs

Usages of the <p> tag should be relatively rare, since we use CSS styles (see below) as much as possible for controlling the margin between elements like lists and tables. Do not use <p> for the first paragraph in a package, class or member documentation, or for the first paragraph after a </ul>, </ol>, </table>, </blockquote>, </pre>, or {@snippet} element. The <p> tag shall be used only for separating a new paragraph from a previous one. In such cases, <p> shall have a matching </p> tag at the paragraph end in order to form valid HTML.

Javadoc CSS

Avoid using HTML attributes other than class as much as possible. Instead, rely on styling. Some HTML tags having a style definition in Apache SIS are:

MathML elements

The source code occasionally uses MathML for formulas that are difficult to render with only Unicode characters. PNG images are not extensively used for formulas because they are difficult to edit after creation, and their content are invisible to search operations (for example when a variable is renamed). For examples of MathML usage in SIS, search for the <math …> XML tag in Java source files. For an introduction to MathML, see:

• MathML learn & use on W3C
• Working with MathML on Wolfram Mathematica

MathML is supported natively in Firefox, Safari and Opera. Internet Explorer users need to install a plugin. Firefox users can optionally install the fonts for Mozilla’s MathML engine for better results. Note that a JavaScript display engine is available for all browsers, but not yet used by SIS.

Compiler warnings

When a local variable in a method has the same name as a field in the enclosing class, some compilers emit a “Local variable hides a field” warning. This warning can be disabled with a @SuppressWarnings annotation. However, by convention Apache SIS applies this annotation only when the local variable should have the same value as the field. Otherwise, the warning should be resolved by renaming a variable. When a code hides a field, it should be a statement such as:

@SuppressWarnings("LocalVariableHidesMemberVariable")
final Foo foo = this.foo;   // May also be `getFoo()`

Those statements may exist for the following reasons:

• this.foo is non-final and the developer wants to make sure that the field is not modified by accident in the method.
• this.foo is a volatile field, therefore should be read only once and cached in the method for performance reasons.
• getFoo() computes the value of this.foo lazily.

Logging

Apache SIS uses the java.util.logging framework. It does not necessarily mean that all SIS users are forced to use this framework, as it is possible to use java.util.logging as an API and have logging redirected to another system. For example, the logging can be redirect to SLF4J by adding the jul-to-slf4j dependency to a project.

The logger names are usually the package name of the class emitting log messages, but not necessarily. In particular, we do not follow this convention if the class is located in an internal package (org.apache.sis.internal.*) since those packages are considered private. In such cases, the logger name should be the package name of the public class invoking the internal method. The reason for that rule is that logger names are considered part of the public API, since developers use them for configuring their logging (verbosity, destination, etc.).

All logging at Level.INFO or above shall be targeted to users or administrators, not to developers. In particular Level.SEVERE shall be reserved for critical errors that compromise the application stability — it shall not be used for exceptions thrown while parsing user data (file or database).