When a spec defines a Model at the top level that is a non-aggretate type (such
as string, number or boolean), it essentially represents an alias for the simple
type. For example, the following spec snippet creates an alias of the boolean
type that for all intents and purposes acts just like a regular boolean.
definitions:
JustABoolean:
type: boolean
This can be modeled in some languages through built-in mechanisms, such as
typedefs in C++. Java, however, just not have a clean way of representing this.
This change introduces an internal mechanism for representing aliases. It
maintains a map in DefaultCodegen that tracks these types of definitions, and
wherever it sees the "JustABoolean" type in the spec, it generates code that
uses the built-in "Boolean" instead.
This functionality currenlty only applies to Java, but could be extended to
other languages later.
The change adds a few examples of this to the fake endpoint spec for testing,
which means all of the samples change as well.
Commit e3d04ee01 (issue #5240) introduced unsafe add/put methods for optional
list/map parameters. This change maintains the spirit of issue #5240 (optional
containers are null by default) while still making add/put calls safe. It does
this by checking for null first and, if so, initializing it with an empty
container.
Also updated the affected samples using the various scripts in bin/.
* set list to null in java if not required
* update petstore sample for java server stub
* update sample for msf4j
* add null assignment to new list/map in java
When a string enumeration has the empty string as one of its available
values, the generated code for many languages is invalid because the
empty string can not be used as an identifier. As with numbers and
symbols, provide a mapping to an English name which can be used as a
replacement. In this case, "empty" for the empty string/empty value.
Signed-off-by: Kevin Locke <kevin@kevinlocke.name>
