* Added support for application/xml content-type for GO language
Issue #7463
* Added test scripts for Go lang "withXml" feature
* Added samples for Go land "withXml" feature.
* "withXml" feature for Go language is only available for client.
* [csharp] ctor params should always be camelCase
After PR #6305, var names defaulted to PascalCase results in constructor
arguments also being PacalCase. Model properties and constructor
arguments have no reason to be the same case, and in fact may cause
issues (`name = name` will result in a compilation error).
This commit forces all constructor params in models to lowerCase.
This is a necessary change, for instance, if client SDK consumers assign
using named args:
var a = new Model(first = "", second = "")
The PacalCase default and update to constructor arg casing will break
existing consumers of the client.
See #7070 for more details and discussion.
* [csharp] Regenerate samples
* [csharp] Remove client models generated from a different spec.
* [csharp] Escape reserved words on camelcase/lowercase lambdas
* [csharp] Regenerate samples
Using the option `taggedUnions` will create a union type for each parent
type instead of extending interfaces. The union types are tagged by using
the discriminator values.
And also:
* Add support for aliases;
* Add support for read-only properties.
* Issue 7440 Added support for returning response in jaxrs-spec interfaces.
* Issue-7386 Updates samples for jaxrs-spec interfaces introduced in issue 7386.
- add new config options:
cabalPackage
Set the cabal package name, which consists of one or more alphanumeric words separated by hyphens
cabalVersion
Set the cabal version number, consisting of a sequence of one or more integers separated by dots
baseModule
Set the base module namespace
requestType
Set the name of the type used to generate requests
configType
Set the name of the type used for configuration
* [tools] Make sed in new.sh more cross-platform
The -r option passed to sed is a GNU sed option for extended regex
evaluation. The -E option evaluates the same option, and is part of the
POSIX standard, meaning this option is available in GNU sed as well as
Apple's BSD variant.
This commit removes the need for users to install gnu-sed on Mac.
* [ktor] Initial ktor (kotlin-server)
This adds a very barebones implementation for a ktor server generator.
This supports metrics and typed locations. All endpoins are stubbed to
return HTTP/1.1 501 Not Implemented.
* [ktor] Initial sample
* [ktor] Adding options for select feature installs
Options available:
* featureAutoHead
* featureConditionalHeaders
* featureHSTS
* featureCORS
* featureCompression
* [ktor] Start of auth functionality
* [ktor] API key auth placeholder
* Add basic support for oauth2 configurations
ktor doesn't seem to explicitly accept oauth flow properties in its
configuration object. This may be a blocker for 'implicit' flow
definitions.
* Added example response objects
* [ktor] Route for apis with bodies, some cleanup
ktor locations are only supported for routes with path/query parameters.
Routes with body or file parameters must be declared with traditional
route api.
This commit also includes lambdas for simplifying processing in
library-based server generator code. As an example, ktor requires
lowercase http methods while spring (a potential future generator)
would require an uppercase such as HttpMethod.GET. It doesn't make sense
to modify these in the operations post-process method because that
format wouldn't be universally desirable.
The lambdas included in the KotlinServerCodegen:
* lowercase: converts all text to lowercase
* uppercase: converts all text to UPPERCASE
* titlecase: converts words (with configurable delim) to Title Case
* indented|indented_8|indented_12|indented_16: these helpers apply the
same desired indent to all lines of an included fragment's text.
* Fix some javadoc issues in lambda classes
* Update kotlin-server-petstore.bat
Change `kotlin` to `kotlin-server`
* Fix javadoc error messages in CI
* Split up model template into partials
* Change models from class to struct.
This fixes issue https://github.com/swagger-api/swagger-codegen/issues/6941.
In this change, we make our Swift4 generated model objects struct instead of class. However, in order to do this, we needed to handle the following edge cases:
* Inheritance and polymorphism (allOf)
* With classes, we use inheritance. So therefore, the parent properties are ONLY on the parent generated class, and the model object which derives from the parent class picks up those properties through inheritance.
* However, structs do not support inheritance. So we simply duplicate the parent allOf properties in the child struct.
* We have to handle the case where the property name on the struct may be different than the property name in the JSON. By default, the Codable protocol assumes that the JSON property name is the same as the struct property name. If they need to be different, then we generate a CodingKeys string enum, which contains the mapping between struct property name and JSON property name.
* additionalProperties. We cannot use the default Codable implementation for the additionalProperties, since it will look for an actual dictionary called "additionalProperties" in the JSON. Therefore, for model objects which have additionalProperties, we must generate our own implementation for the Decodable and Encodable protocols.
I have run ./bin/swift4-all.sh and ./bin/swift4-test.sh to re-generate all of the sources, and I have verified that the generated code in samples/clients/test/swift4/default builds and the unit tests pass.
* Update VERSION in .swagger-codegen
* Update generated code for swift4-test schema
* [csharp] Support composition on toJson
Previous implementation assumed specification only supports polymorphic
associations (via discrimator), although the code didn't seem to be
setup correctly for that in the first place. That is, the parent object
must define the discriminator (see
https://github.com/OAI/OpenAPI-Specification/blob/master/versions/2.0.md#models-with-polymorphism-support),
so NOT HAS parent AND HAS discriminator doesn't make sense.
From a C# perspective, base classes should have the method marked
virtual and derived classes should override the method. This supports
both composition and polymorphic definitions.
* [csharp] this.Configuration in api template
Unprefixed Configuration property access leads to ambiguous references
when spec defines a Configuration model.
* [csharp] Models/properties support nested arrays
Previous implementation didn't support multiple levels of array with
array items as OpenAPI spec supports. This means an object defined as
type: array with items = type: array|items=double (which is common in
GIS) would not be possible.
This implementation assumes generics in the nested type definitions, so
the above would generate List<List<double?>> for model parent types as
well as property type declarations.
* [csharp] Regenerate integration test sample
* [csharp] Set "Client" case sensitive as reserved
* [csharp] Regenerate security sample
* [csharp] Regenerate samples
* [csharp] Support composition on toJson
Previous implementation assumed specification only supports polymorphic
associations (via discrimator), although the code didn't seem to be
setup correctly for that in the first place. That is, the parent object
must define the discriminator (see
https://github.com/OAI/OpenAPI-Specification/blob/master/versions/2.0.md#models-with-polymorphism-support),
so NOT HAS parent AND HAS discriminator doesn't make sense.
From a C# perspective, base classes should have the method marked
virtual and derived classes should override the method. This supports
both composition and polymorphic definitions.
* [csharp] Regenerate integration test files
* [csharp] Regenerate samples
* [csharp] Regenerate security sample
* [scala] Escape reserved words, support Array[Byte]
Previously, Array[Byte] was compiling to ArrayByte. This provides a type
mapping to output the correct type.
This also escapes reserved words with grave accents, as is most common
in Scala. Escaping with an underscore prefix breaks serialization (in
Jackson, for example) unless templates are modified manually. Escaping
using grave accent should unblock most serializers from requiring
template modifications.
* [scala] Regenerate integration test outputs
* [scala] Regenerate samples
* [scala] Remove unused imports in related codegen files
* [scala] Fix default values in scala client
This uses consistent logic for optional types with default values in the
scala client. Also, uses Option(default) instead of Some(default) to
guard against people defining defaultValue = null. Option(null) becomes
None while Some(null) defines a null value explicitly and will break
maplike operations.
* [scala] Regenerate client sample
* [scala] Add missing json4s import, which will be added by another PR but allows current samples to generate
* [scala] Include integration tests for required attributes support
* [scala] Support string types with formats
This adds support for better support of type=string and
format={date,date-time,binary,byte}. Previously, binary and byte were
inconsistently defined as strings rather than byte arrays, while
date/date-time were parsing default values into formats that did not
match OpenAPI/Swagger 2.0 specifications for full-date and date-time.
We may want to consider pulling in json4s-ext to support wider date
formats and moving to date=LocalDate and date-time=ZonedDateTime.
This will have breaking changes for consumers expecting binary/byte to
be strings rather than byte arrays.
* [scala] Unique parameter names in integration test, to avoid seemingly conflicting names
* [scala] Regenerate client sample
* Sort file listings in AssertFile.java
Per File#list() javadocs:
There is no guarantee that the name strings in the resulting array
will appear in any specific order; they are not, in particular,
guaranteed to appear in alphabetical order.
I'm unable to repro directory listing failures on OS X High Sierra or
Ubuntu 16.04 under Parallels, so it's not clear to me if listing order
is indeterminate per-platform or the behavior is just not defined and
up to the platform's installed runtime. Sorting the array of strings
prior to comparison should resolve this issue on every platform/runtime.
* [scala] exclude api tests for integration test gen script
Script should match options in the integration test class
* [scala] Temporarily disable client integration tests
CI doesn't seem to pick up template changes in integration tests.
Disabling scala client integration tests, pending investigation of the
issue.
* [scala] Remove redundant json4s import
* [scala] Regenerate integration test
* [scala] Regenerate sample
* [haskell-http-client] handle */* mimetype correctly
* [haskell-http-client] generate separate api modules, for each child api
* [haskell-http-client] add cliOption "allowNonUniqueOperationIds"
When cli option "allowNonUniqueOperationIds" is true,
allows *different* API modules to contain the same operationId,
and then each API must be imported qualified.
* MMORCH-428: Export the Authorization struct
This is needed so that code can check that the version of Authorization
in the auto-generated code is the same as the version it is using. If
the versions are not exactly the same then the lookup into the TypeMap
will not work.
* Add Rust as a supported language for client and server.
Clarify that there are two Rust client implementations, and one Rust server implementation.
* Percent-encode path and query parameters in client URLs
Fixes#122
Also don't include a question mark at the end of the path when there are no query paramters.
Fixes#121
* Rust2 client: add --host and --port parameters to example client.
Allows the example command-line client to override the default host and port.
* Extract default host and port from Swagger file.
* Derive 'Eq' and 'Ord' on enums
* Rust2: improve server code structure.
server.rs (main.rs) - main entry point for binary; starts the web server and points it at the server code within the library.
server_lib/mod.rs (lib.rs) - root of library; creates the server.
server_lib/server.rs (server.rs) - actual server code
The old server_lib/mod.rs is now server_lib/server.rs; server_lib/mod.rs is new.
This structure is easy to map onto a server implementation; unfortunately we can't get it exactly right here because of the limitations of cargo's examples/ folder.
* Rust2: Explain fully how to use the example code in your project.
* Added plaintext support
* Linting
* MGJ Markups
* Move errant file to 'rust-server'
Rust2 was the old name
* Add license and description to Cargo.toml
For rust-server, getting them from the swagger
* Valid Java doesn't have uninitialised variables...
* Default license to "Unlicense"
* Set license to "Unlicense" for generated crates
* Do not apply content-type header if no request body
* Set content-type header if there are form params
* Add tests
* Run update script
* Regenerated sample clients
* Initial Commit
* Removed unneeded class, Updated names as per naming conventions, Removed licensing info from generated code
* Moved samples from client folder to server
* Fixed naming issue with scripts in the bin
* improved code format for api
* fixed compile issue and improved formatting
* fixed compile issue and improved formatting
* [csharp] Convert "false" properties to booleans
It appears as though "false" strings in additionalProperties are no
longer treated as false booleans. This may be an issue elsewhere, but a
simple fix is to always explicitly set the boolean value in a generator
class back to the additionalProperties map to convert boolean Strings to
boolean Objects.
* [nancyfx] Clean up async default option handling
* [nancyfx] Include asyncServer=false in sample script
* [csharp] Regenerate samples
* [csharp] Resolve .net 4 generation issues
Some functionality is missing from .NET 4.0, such as IReadonlyDictionary
and Type.GetTypeInfo().
This commit resolves compilation of generated .NET 4.0 code, requiring
no conditional versioning of Newtonsoft.Json.
* [csharp] Regenerate .net 4.0 sample
* [csharp] Resolve .NET 4.0 sample compile
Sample build.sh wasn't accounting for targeting different FCL correctly.
That is, when passing "net40" to the -sdk option, it would use the
default -sdk:4 and -langversion:6. These don't necessarily match with
what is installed on a machine with only .NET 4.0 (which is our targeted
use case here).
To resolve, we need to define another version-specific value for passing
to the mcs -sdk option (see man mcs for details).
This option currently isn't overridable in the client codegen class.
Also, langversion is set specifically to the version of C# available to
the targeted SDK version. If there is need, we may extend this to
something like:
langversion=${MCS_LANG_VERSION:-6}
To allow users to run as:
env MCS_LANG_VERSION=5 sh build.sh
I haven't done this because I doubt there's much of a use case via this
script. I'm assuming most consumers will build via IDE or MSBuild.
* [csharp] Revert bin/csharp-petstore.sh to 3.5
* [csharp] Regenerate .NET 3.5 sample
* [csharp] Resolve nuget issue with existing files
* [csharp] Update -all.sh, regenerate samples
* master: (101 commits)
[Swift4] Allow for custom dateformatter to be used (#6672)
[haskell-http-client] fix bug when generating models-only (#6931)
fix typo: crediential => credential
minor typo fix
[csharp] fix enum serialization of first value (#6873)
[PHP] Improve docs and README (#6935)
Binary mode for file deserialization in python (#6936)
add python tornado test to travis
[Python/tornado] add integration tests and fix bugs (#6925)
Fix PHP passes response body to ApiException (#6923)
[TypeScript][Node] Resolve TS2532 error (#6932)
skip "all" shell script
minor formatting change
Fixes Issue #6841, Map for accessing additionalProperties is generated. (#6886)
add tsloughter as owner erlang
WIP: initial commit for Erlang client generator (#6502)
add back php client test
Switch Travis image from MacOS to Linux (#6937)
add link to ebook
[Scala] Default case class Option types to None for non-required fields (#6790)
...
* [csharp] Treat enum models consistently
C# works differently from most languages in that enums are not
considered objects. This means default(EnumType) will choose a default
of the first enum option. This isn't desirable because it breaks the
required = false functionality of swagger specs, which defines a
property which isn't required to exist in the message body.
Rather than force consumers to use enum values such as UNSPECIFIED, UNKNOWN,
NOT_SET, etc... we can treat enums as primitives. This means any
non-required enum will become Nullable<EnumType> regardless of whether
it is defined as an inline enum or a referenced enum model.
* Categorizing C# integration test for enums as general
* [csharp] Remove enum-ref integration test
* [csharp] Clean up general enum support integration test, validate different enum usage cases.
* [csharp][all] Assign one-based int to string enums
The EmitDefaultValue=false for string based enums will prevent the first
enum value from being serialized, because as 0 it is considered the
default.
This commit assigns an explicit numerical value to all non-integer
enums. This assignment has no effect on the
serialization/deserialization values, and only assigns the compiled
integer.
NOTE: This will have an effect of requiring recompilation of any code
that references the client/server models. This is because:
public enum Pet { Available }
Source files referencing Pet.Available as defined above will have a
constant 0 in place of the enum value.
public enum Pet { Available = 1 }
Source files referencing Pet.Available as defined above will have a
constant 1 in place of the enum value.
After compilation, Pet.Available in both instances lose their semantic
values and refer to the byte representation of their integral values.
For more info, see
https://docs.microsoft.com/en-us/dotnet/csharp/language-reference/keywords/enum
* [csharp] Treat enum models consistently
C# works differently from most languages in that enums are not
considered objects. This means default(EnumType) will choose a default
of the first enum option. This isn't desirable because it breaks the
required = false functionality of swagger specs, which defines a
property which isn't required to exist in the message body.
Rather than force consumers to use enum values such as UNSPECIFIED, UNKNOWN,
NOT_SET, etc... we can treat enums as primitives. This means any
non-required enum will become Nullable<EnumType> regardless of whether
it is defined as an inline enum or a referenced enum model.
* Categorizing C# integration test for enums as general
* [csharp] Remove enum-ref integration test
* [csharp] Clean up general enum support integration test, validate different enum usage cases.
Enums defines as ref models have a different object structure
(CodegenModel) than those defined as inner enums (CodegenProperty). To
make these look as similar as possible, we walk all ref model enums and
reassign enumVars with the same properties inherited from the top level
CodegenProperty so CodegenModel enums can use the same templates as
inner enums.
* [kotlin] Nested enum naming fix, and naming options via CLI
* [kotlin] Add option test for enum property naming
* [kotlin] Escape all reserved/keywords for enums
