OpenMRS tries to follow a common coding style in order to make our code base robust: more readable and more reliable (if the same code is created by three different developers, it should look and read very similar). Our style conventions are designed to promote consistency and good coding practices, while reducing common errors. If you feel that our coding styles could be improved (by changing, adding, or removing something), please bring your comments to the Developers Mailing List.

We use a combination of tools (checkstyle, PMD, FindBugs, ...) to write down our coding conventions in a set of rules. These rules are picked up by services (provided by our own infrastructure like SonarQube or cloud providers like codacy) so that any addition of new code (via a commit or pull request on Github) is checked for violations of these rules. Having the rules in simple config files together with the source code enable any developer to run the same checks locally.

Tools

The following tools are used to ease code review, help new developers get used to our coding style and maintain code quality.

Checkstyle

We use checkstyle for static code analysis. checkstyle defines rules http://checkstyle.sourceforge.net/checks.html which when violated in the code will result in errors or warnings. The checkstyle rules we want our code to adhere to are defined in the openmrs-core repostitory's checkstyle.xml

Manual Run

You can run checkstyle locally using the maven checkstyle plugin with the following command

mvn clean compile jxr:aggregate checkstyle:checkstyle-aggregate

This will generate an html report at

target/site/checkstyle-aggregate.html

(the path is relative to the root of the openmrs-core repository)

The report will show you all the violations for each java class with their severity, the name of the rule and the location (the jxr portion in the above command enables you to navigate to the location of the violation in the java class). If you want to know more about a specific rule go to http://checkstyle.sourceforge.net/checks.html

PMD

PMD is used to check for issues like unused, unnecessary or error prone code. The PMD rules we want our code to adhere to are defined in the openmrs-core repostitory's ruleset.xml

Manual Run

Install PMD locally and try

./pmd-bin-5.5.4/bin/run.sh pmd -d openmrs/openmrs-core/api/src/main/java -f text -R openmrs/openmrs-core/ruleset.xml -version 1.8 -language java

Adjust the version and location of the PMD you installed and refer to the PDM documentation to see the available command line options.

FindBugs

FindBugs is another tool we use to check the code for issues. The FindBugs rules are not yet in the repository on Github, at the moment you can find them at the SonarQube.

Manual Run

At the moment you cannot check the code against the FindBugs rules locally using maven as with checkstyle. We are working on it (smile)

Automation

Codacy & Github Integration

We have set up integration with codacy so that our rules will be checked on every pull request. Whenever you make a pull request your code will be commented by codacy and if its clean the check will pass and if not show you what parts of your code violated which rules. You can see the current status of the openmrs-core at https://www.codacy.com/app/openmrs/openmrs-core/dashboard

SonarQube

SonarQube checks the openmrs-core code once a day. You can find the issues it found here

Read FindBugs to see where the rules come from.

Limitations & Work In Progress

We aim to bring all our rules (FindBugs, ESLint) into the openmrs-core repository and to let codacy use both set of rules to check the code. Codacy is currently configured to read our checkstyle and PMD rules from the repository to check Java code that goes into it.

Other languages such as javascript, ruby, ... might be checked as well but we do not yet have formalized these rules into config files like ESLint or JSHint. If you want to do that go ahead!

FindBugs rules will not be checked by codacy since that is a feature only available to enterprise users. If you know another cloud provider that satisfies our needs and supports FindBugs please tell us!

The topic is discussed here:

How-To Write Code According to the Conventions

Make sure you have followed the How-To Setup And Use Your IDE so that your code changes will automatically be formatted according to the OpenMRS style.

Unit Tests

Conventions

The following conventions are applied to code in openmrs-core but we encourage module authors to adopt them as well.

File template

Attribution

 

(NOTE: this attribution-free coding policy took effect after version 1.1, so you may see some attribution in the existing code; we are gradually removing attribution from our code base)

Code Style

Imports

Make sure you have followed the How-To Setup And Use Your ID which will prevent you from running in two the following issues.

Remove unused imports

Imports that are not used should be removed, remember that you introduce dependencies that are not needed!

Do not use wildcard imports

If you are importing multiple classes from a package, list them all individually, like in the following code block, since this makes it clear exactly what imported classes are in use in any file:

import java.util.ArrayList;
import java.util.List;
import java.util.Random;

Do not do the following:

import java.util.*;

Please check your IDE settings it is often the IDE which merges several imports into a wildcard import.

Naming Conventions

Strings

Concatenation

Use of StringBuffer in place of String Concatenation. Use the StringBuffer if you want to concatenate a number of string literals especially when this concatenation is not being done in a single line. e.g.
Don't write like String s = ""; s+= "Hey!"; s+'=" "; s+="Bye!" ;
Always write StringBuffer sb = new StringBuffer(); sb.append(Hey!); sb.append(" "); sb.append("Bye!"); String s = sb.toString();

Use of StringUtils

Do not comment out code

Do not leave commented out code in the source files. Just remove it. Others will think this has been left there for a purpose and thus will probably not dare to delete it which will leave this commented out code hanging around forever. There are some exceptions of course, such as: "When I'm pretty sure it needs re-enabled at some point, and to make sure I don't forget why that is, I leave a clear TODO comment, e.g. "TODO this needs re-enabled when TRUNK-XXXX is fixed". Most IDEs are good at parsing out all TODO statements." Rowan Seymour

Logging

Background

OpenMRS uses the logging facade https://www.slf4j.org/ which in turn delegates to the log4j as the logging implementation. If you want some background info on these libraries and their relation to Apache commons-logging and spring read this http://docs.spring.io/spring/docs/4.1.4.RELEASE/spring-framework-reference/htmlsingle/#overview-not-using-commons-logging

How To Log

If you want to log you should rely on the slf4j. The following shows a typical usage borrowed from their API documentation

 import org.slf4j.Logger;
 import org.slf4j.LoggerFactory;

 public class Wombat {

   protected final Logger log = LoggerFactory.getLogger(Wombat.class);
   Integer t;
   Integer oldT;

   public void setTemperature(Integer temperature) {
     oldT = t;
     t = temperature;
     log.debug("Temperature set to {}. Old temperature was {}.", t, oldT);
     if(temperature.intValue() > 50) {
       log.info("Temperature has risen above 50 degrees.");
     }
   }
 }

How To (Not) Log

You might need to construct a message for logging and want to prevent this construction from happening if the debug log level is not enabled. You could then come up with a pattern like this

if(log.isDebugEnabled()) {
  log.debug("Entry number: " + i + " is " + String.valueOf(entry[i]));
}

which you should not use even if you see it in existing code (if you see this clean it up).

Instead you should use parametrized messages:

log.debug("The entry is {}.", entry);

Why? Read https://www.slf4j.org/faq.html#logging_performance for the answer.

Code Templates

Add the OpenMRS code templates to your IDE (Eclipse or IntelliJ) which will save you typing and help you stick to the above conventions.

More Conventions

Do not print out

Modifier order

Modifiers should follow the order suggested by the Java Language specification, sections 8.1.1, 8.3.1, 8.4.3 and 9.4:

  1. public
  2. protected
  3. private
  4. abstract
  5. default
  6. static
  7. final
  8. transient
  9. volatile
  10. synchronized
  11. native
  12. strictfp

We enforce this via Checkstyle rule ModifierOrder

Others

Documentation

Exception Handling

General

Creating Custom Exceptions

Create an intuitive hierarchy of exceptions within the package they belong, only creating new Exception classes for when the need to branch on different types of exception within code/webapp are needed (add as needed instead of making extra classes "just in case" for every possible variation).

For example:

org.openmrs.api.APIException
org.openmrs.api.PatientIdentifierException extends APIException
org.openmrs.api.MissingPatientIdentifierException extends PatientIdentifierException

and later add:

org.openmrs.api.DuplicatePatientIdentifierException extends PatientIdentiferException

when we realize the webapp needs to distinguish duplicates from invalid identifier exceptions.

Patching Third Party Libraries

New Public Methods and Classes

All new public methods and classes should have a @since annotation for the version when they were introduced. If the class is new, then it is enough to just put it at the class level instead of each method. Something like: @since 2.0

Deprecation

Security

Avoiding XSS scripting