Major Changes From Version 2 to 3
New features
There are a lot of new features, see the home page of this documentation for more information.
Most notably new features:
- Better boolean support.
- Support for string operations and functions.
- Support for date-time and duration values and operations.
- Support for null values.
- Support for multidimensional arrays.
- Support for nestable structures.
- Structures and arrays can be combined to work with arbitrary data structures.
- New data access support, connecting expressions with you data made easy.
- New configuration concept.
License
The license has changed from MIT to Apache 2.0. This should not be a problem, as both licenses are very similar. The Apache 2.0 license though is clearer about certain constraints to some use cases.
New Java version
EvalEx 3 now requires at least Java 11 to run and compile.
New package, group id and repository home.
EvalEx has moved from a personal repository to an organizational repository. This decouples the product from a single person and allows more and better control over the product.
Complete rewrite
EvalEx 3 is a complete rewrite. After 10 years of adding features and trying to stay backward compatible, I felt that this is the time for a big cut. Trying to stay backward compatible has introduced several code constructs, that didn’t feel good.
This cut has direct impact to the integration of EvalEx in an existing application. Though I believe that the new version has much better integration possibilities (e.g. the data access interface, the separate configuration object), existing integrations need some refactorings.
Depending on how you used EvalEx until now, this may not need very much changes, e.g.:
- Change the dependency coordinates of EvalEx to the new group id.
- Change the package imports.
- Adopt the new method and class names.
- Use the new return value (EvaluationValue instead of BigDecimal).
When you were using custom functions or operators, or own configurations, then you may need a bit more work. Though even these migrations are straight forward and pretty easy.
Here are the main changes, when it comes to integration:
New configuration concept
Configurations are now separated from the expression. You can create a configuration once and then re-use it with your expressions. All the heavy constructor work is now in the configuration, making the instantiation of an expression much faster.
New default MathContext
The new default MathContext has a precision of 68 with a HALF_EVEN rounding mode, compared to a precision of 7 in EvalEx 2. This is a major leap forward. Low precision has been a constant issue with EvalEx 2.
Moved from RPN to AST
EvalEx 2 used RPN (Reverse Polish Notation) for evaluation. The parser transformed the infix notation to an RPN notation and used this representation for evaluation.
Now the parser transforms the infix notation to an AST (Abstract Syntax Tree), allowing several improvements and new features. The main reason for the switch was the lazy evaluation needed for the IF function. With the use of an AST, this was easy to achieve.
If you were using the getRPN() function in EvalEx 2, then the new getAllASTNodes() or getAbstractSyntaxTree() methods may be useful for you. If you still need RPN support, it should be possible to create an RPN notation out of the AST with a bit of effort.
New return value type
EvalEx 2 evaluation always returned a BigDecimal. With the new support of strings, arrays and structures, this was not suitable anymore. The new EvaluationValue return type for an evaluation has a type member that defines what data type it holds (boolean, number, string, array, structure, or an AST). It also has some convenience methods to convert the values, e.g. number to string.
New data access concept
EvalEx 2 had an own storage for variable and constant values, that had to be filled with the required values before evaluation. This is also the default with EvalEx 3, so no change to existing integrations is required.
But the new concept allows additionally to configure a custom data access interface, which could be backed by any kind of storage. See chapter Data Access for details.
Custom Operators
Adding custom operators is now easier, but has changed significantly. See chapter Custom Operators for details.
Custom functions
Adding custom functions is now easier, but has changed significantly. See chapter Custom Functions for details.