Skip to content

Latest commit

 

History

History
87 lines (70 loc) · 4.79 KB

NOTES.md

File metadata and controls

87 lines (70 loc) · 4.79 KB
Raw

Notes

Recent directory structure change

Due to a recent commit - #515 Merge tests and pom and code - the structure of the project has changed from a flat directory containing all of the Java files to a directory structure that includes unit tests and several tools used to build the project jar and run the unit tests. If you have difficulty using the new structure, please open an issue so we can work through it.

Implementation notes

Numeric types in this package comply with ECMA-404: The JSON Data Interchange Format and RFC 8259: The JavaScript Object Notation (JSON) Data Interchange Format. This package fully supports Integer, Long, and Double Java types. Partial support for BigInteger and BigDecimal values in JSONObject and JSONArray objects is provided in the form of get(), opt(), and put() API methods.

Although 1.6 compatibility is currently supported, it is not a project goal and might be removed in some future release.

In compliance with RFC8259 page 10 section 9, the parser is more lax with what is valid JSON then the Generator. For Example, the tab character (U+0009) is allowed when reading JSON Text strings, but when output by the Generator, the tab is properly converted to \t in the string. Other instances may occur where reading invalid JSON text does not cause an error to be generated. Malformed JSON Texts such as missing end " (quote) on strings or invalid number formats (1.2e6.3) will cause errors as such documents can not be read reliably.

Some notable exceptions that the JSON Parser in this library accepts are:

  • Unquoted keys { key: "value" }
  • Unquoted values { "key": value }
  • Unescaped literals like "tab" in string values { "key": "value with an unescaped tab" }
  • Numbers out of range for Double or Long are parsed as strings

Recent pull requests added a new method putAll on the JSONArray. The putAll method works similarly to other put methods in that it does not call JSONObject.wrap for items added. This can lead to inconsistent object representation in JSONArray structures.

For example, code like this will create a mixed JSONArray, some items wrapped, others not:

SomeBean[] myArr = new SomeBean[]{ new SomeBean(1), new SomeBean(2) };
// these will be wrapped
JSONArray jArr = new JSONArray(myArr);
// these will not be wrapped
jArr.putAll(new SomeBean[]{ new SomeBean(3), new SomeBean(4) });

For structure consistency, it would be recommended that the above code is changed to look like 1 of 2 ways.

Option 1:

SomeBean[] myArr = new SomeBean[]{ new SomeBean(1), new SomeBean(2) };
JSONArray jArr = new JSONArray();
// these will not be wrapped
jArr.putAll(myArr);
// these will not be wrapped
jArr.putAll(new SomeBean[]{ new SomeBean(3), new SomeBean(4) });
// our jArr is now consistent.

Option 2:

SomeBean[] myArr = new SomeBean[]{ new SomeBean(1), new SomeBean(2) };
// these will be wrapped
JSONArray jArr = new JSONArray(myArr);
// these will be wrapped
jArr.putAll(new JSONArray(new SomeBean[]{ new SomeBean(3), new SomeBean(4) }));
// our jArr is now consistent.

Unit Test Conventions

Test filenames should consist of the name of the module being tested, with the suffix "Test". For example, Cookie.java is tested by CookieTest.java.

The fundamental issues with JSON-Java testing are:

  • JSONObjects are unordered, making simple string comparison ineffective.
  • Comparisons via equals() is not currently supported. Neither JSONArray nor JSONObject override hashCode() or equals(), so comparison defaults to the Object equals(), which is not useful.
  • Access to the JSONArray and JSONObject internal containers for comparison is not currently available.

General issues with unit testing are:

  • Just writing tests to make coverage goals tends to result in poor tests.
  • Unit tests are a form of documentation - how a given method works is demonstrated by the test. So for a code reviewer or future developer looking at code a good test helps explain how a function is supposed to work according to the original author. This can be difficult if you are not the original developer.
  • It is difficult to evaluate unit tests in a vacuum. You also need to see the code being tested to understand if a test is good.
  • Without unit tests, it is hard to feel confident about the quality of the code, especially when fixing bugs or refactoring. Good tests prevent regressions and keep the intent of the code correct.
  • If you have unit test results along with pull requests, the reviewer has an easier time understanding your code and determining if it works as intended.