[php][DefaultCodegen.java] Fix invalid enum const names

[githubERIK]

@jebentier, @dkarlovi, @mandrean, @jfastnacht, @ackintosh, @ybelenko, @renepardon

Fix#1475

[githubERIK]

How is it possible that some enum 1 is transformed to ENUM_INTEGER_1 while some numeric enums just stay the same (this pull request wants (EDIT: wanted) to add the prefix "N" to these cases)?

One potential idea: enums that are pure schema objects should get the name of the schema as their prefix.

[githubERIK]

Reverted add-N-prefix changes with aa36b80 and d07efac.

The new approach, however, does not affect the clients (the generated PHP clients still have const 0=0;) which means that the changed method toEnumVarName is not called properly.

[githubERIK]

It seems to be a good idea to debug the generated jar and start with a breakpoint in DefaultCodegen and then perform "Step Into" when "importPath→OpenAPI\Client\Model.EnumTest" appears.

and then see what happens with outerEnumInteger

Enum_Test:
type: objectrequired:
- enum_string_requiredproperties:
enum_string:
type: stringenum:
- UPPER
- lower
- ''enum_string_required:
type: stringenum:
- UPPER
- lower
- ''enum_integer:
type: integerformat: int32enum:
- 1
- -1enum_number:
type: numberformat: doubleenum:
- 1.1
- -1.2outerEnum:
$ref: '#/components/schemas/OuterEnum'outerEnumInteger:
$ref: '#/components/schemas/OuterEnumInteger'outerEnumDefaultValue:
$ref: '#/components/schemas/OuterEnumDefaultValue'outerEnumIntegerDefaultValue:
$ref: '#/components/schemas/OuterEnumIntegerDefaultValue'

d07efac leads to

at some point.

But in the end

class OuterEnumInteger
{
/** * Possible values of this enum */const0 = 0;
const1 = 1;
const2 = 2;
/** * Gets allowable values of the enum * @return string[] */publicstaticfunctiongetAllowableEnumValues()
{
return [
self::0,
self::1,
self::2,
];
}
}

is generated.

Probably because after OuterEnumInteger is processed as part of EnumTest it is also processed later separately:

.

[githubERIK]

The final approach: change postProcessModelsEnum in DefaultCodegen.java to prevent adding enum variable names that are numbers. If number is detected, add a prefix NUMBER_ (see 46b9927).

Generated all samples by executing shell scripts in bin and bin/openapi3 folders using a script generate_all_sample_clients.sh.

[ybelenko]

Is this PR is still relevant or it should be closed?

[asabramo]

Hi guys,
I see this issue was reported 2 years ago.
When I use the latest:
swagger-codegen-cli-3.0.20.jar generate -l php -i api.yaml -o generated_api

I'm still seeing this issue, which means I can't use the files as they are generated - I have to add another step manually, which beats the purpose.

Oddly enough, when I use the version that the Jetbrains plugin retrieves automatically (https://repo1.maven.org/maven2/io/swagger/codegen/v3/swagger-codegen-cli/3.0.3/swagger-codegen-cli-3.0.3.jar) , the enum problem doesn't exist (but then there's another bug - BigDecimal instead of float).

Is there a version where both issues are fixed?
Is there any workaround for the enum issue?
Thanks,
Assi

[ybelenko]

@asabramo It seems that you're talking about Swagger Codegen which is different project. More information you can read at: https://github.com/OpenAPITools/openapi-generator/blob/master/docs/qna.md

About bug itself, I don't know whether it fixed or not. Maybe we should add part of the spec from related issue to samples and check PHP syntax of generated models with phplint package during CI.

[githubERIK]

@jebentier, @dkarlovi, @mandrean, @jfastnacht, @ackintosh, @ybelenko, @renepardon

Fix#1475

[githubERIK]

How is it possible that some enum 1 is transformed to ENUM_INTEGER_1 while some numeric enums just stay the same (this pull request wants (EDIT: wanted) to add the prefix "N" to these cases)?

One potential idea: enums that are pure schema objects should get the name of the schema as their prefix.

[githubERIK]

Reverted add-N-prefix changes with aa36b80 and d07efac.

The new approach, however, does not affect the clients (the generated PHP clients still have const 0=0;) which means that the changed method toEnumVarName is not called properly.

[githubERIK]

It seems to be a good idea to debug the generated jar and start with a breakpoint in DefaultCodegen and then perform "Step Into" when "importPath→OpenAPI\Client\Model.EnumTest" appears.

and then see what happens with outerEnumInteger

Enum_Test:
type: objectrequired:
- enum_string_requiredproperties:
enum_string:
type: stringenum:
- UPPER
- lower
- ''enum_string_required:
type: stringenum:
- UPPER
- lower
- ''enum_integer:
type: integerformat: int32enum:
- 1
- -1enum_number:
type: numberformat: doubleenum:
- 1.1
- -1.2outerEnum:
$ref: '#/components/schemas/OuterEnum'outerEnumInteger:
$ref: '#/components/schemas/OuterEnumInteger'outerEnumDefaultValue:
$ref: '#/components/schemas/OuterEnumDefaultValue'outerEnumIntegerDefaultValue:
$ref: '#/components/schemas/OuterEnumIntegerDefaultValue'

d07efac leads to

at some point.

But in the end

class OuterEnumInteger
{
/** * Possible values of this enum */const0 = 0;
const1 = 1;
const2 = 2;
/** * Gets allowable values of the enum * @return string[] */publicstaticfunctiongetAllowableEnumValues()
{
return [
self::0,
self::1,
self::2,
];
}
}

is generated.

Probably because after OuterEnumInteger is processed as part of EnumTest it is also processed later separately:

.

[githubERIK]

The final approach: change postProcessModelsEnum in DefaultCodegen.java to prevent adding enum variable names that are numbers. If number is detected, add a prefix NUMBER_ (see 46b9927).

Generated all samples by executing shell scripts in bin and bin/openapi3 folders using a script generate_all_sample_clients.sh.

[ybelenko]

Is this PR is still relevant or it should be closed?

[asabramo]

Hi guys,
I see this issue was reported 2 years ago.
When I use the latest:
swagger-codegen-cli-3.0.20.jar generate -l php -i api.yaml -o generated_api

I'm still seeing this issue, which means I can't use the files as they are generated - I have to add another step manually, which beats the purpose.

Oddly enough, when I use the version that the Jetbrains plugin retrieves automatically (https://repo1.maven.org/maven2/io/swagger/codegen/v3/swagger-codegen-cli/3.0.3/swagger-codegen-cli-3.0.3.jar) , the enum problem doesn't exist (but then there's another bug - BigDecimal instead of float).

Is there a version where both issues are fixed?
Is there any workaround for the enum issue?
Thanks,
Assi

[ybelenko]

@asabramo It seems that you're talking about Swagger Codegen which is different project. More information you can read at: https://github.com/OpenAPITools/openapi-generator/blob/master/docs/qna.md

About bug itself, I don't know whether it fixed or not. Maybe we should add part of the spec from related issue to samples and check PHP syntax of generated models with phplint package during CI.

[githubERIK]

@jebentier, @dkarlovi, @mandrean, @jfastnacht, @ackintosh, @ybelenko, @renepardon

Fix#1475

[githubERIK]

How is it possible that some enum 1 is transformed to ENUM_INTEGER_1 while some numeric enums just stay the same (this pull request wants (EDIT: wanted) to add the prefix "N" to these cases)?

One potential idea: enums that are pure schema objects should get the name of the schema as their prefix.

[githubERIK]

Reverted add-N-prefix changes with aa36b80 and d07efac.

The new approach, however, does not affect the clients (the generated PHP clients still have const 0=0;) which means that the changed method toEnumVarName is not called properly.

[githubERIK]

It seems to be a good idea to debug the generated jar and start with a breakpoint in DefaultCodegen and then perform "Step Into" when "importPath→OpenAPI\Client\Model.EnumTest" appears.

and then see what happens with outerEnumInteger

Enum_Test:
type: objectrequired:
- enum_string_requiredproperties:
enum_string:
type: stringenum:
- UPPER
- lower
- ''enum_string_required:
type: stringenum:
- UPPER
- lower
- ''enum_integer:
type: integerformat: int32enum:
- 1
- -1enum_number:
type: numberformat: doubleenum:
- 1.1
- -1.2outerEnum:
$ref: '#/components/schemas/OuterEnum'outerEnumInteger:
$ref: '#/components/schemas/OuterEnumInteger'outerEnumDefaultValue:
$ref: '#/components/schemas/OuterEnumDefaultValue'outerEnumIntegerDefaultValue:
$ref: '#/components/schemas/OuterEnumIntegerDefaultValue'

d07efac leads to

at some point.

But in the end

class OuterEnumInteger
{
/** * Possible values of this enum */const0 = 0;
const1 = 1;
const2 = 2;
/** * Gets allowable values of the enum * @return string[] */publicstaticfunctiongetAllowableEnumValues()
{
return [
self::0,
self::1,
self::2,
];
}
}

is generated.

Probably because after OuterEnumInteger is processed as part of EnumTest it is also processed later separately:

.

[githubERIK]

The final approach: change postProcessModelsEnum in DefaultCodegen.java to prevent adding enum variable names that are numbers. If number is detected, add a prefix NUMBER_ (see 46b9927).

Generated all samples by executing shell scripts in bin and bin/openapi3 folders using a script generate_all_sample_clients.sh.

[ybelenko]

Is this PR is still relevant or it should be closed?

[asabramo]

Hi guys,
I see this issue was reported 2 years ago.
When I use the latest:
swagger-codegen-cli-3.0.20.jar generate -l php -i api.yaml -o generated_api

I'm still seeing this issue, which means I can't use the files as they are generated - I have to add another step manually, which beats the purpose.

Oddly enough, when I use the version that the Jetbrains plugin retrieves automatically (https://repo1.maven.org/maven2/io/swagger/codegen/v3/swagger-codegen-cli/3.0.3/swagger-codegen-cli-3.0.3.jar) , the enum problem doesn't exist (but then there's another bug - BigDecimal instead of float).

Is there a version where both issues are fixed?
Is there any workaround for the enum issue?
Thanks,
Assi

[ybelenko]

@asabramo It seems that you're talking about Swagger Codegen which is different project. More information you can read at: https://github.com/OpenAPITools/openapi-generator/blob/master/docs/qna.md

About bug itself, I don't know whether it fixed or not. Maybe we should add part of the spec from related issue to samples and check PHP syntax of generated models with phplint package during CI.

[githubERIK]

@jebentier, @dkarlovi, @mandrean, @jfastnacht, @ackintosh, @ybelenko, @renepardon

Fix#1475

[githubERIK]

How is it possible that some enum 1 is transformed to ENUM_INTEGER_1 while some numeric enums just stay the same (this pull request wants (EDIT: wanted) to add the prefix "N" to these cases)?

One potential idea: enums that are pure schema objects should get the name of the schema as their prefix.

[githubERIK]

Reverted add-N-prefix changes with aa36b80 and d07efac.

The new approach, however, does not affect the clients (the generated PHP clients still have const 0=0;) which means that the changed method toEnumVarName is not called properly.

[githubERIK]

It seems to be a good idea to debug the generated jar and start with a breakpoint in DefaultCodegen and then perform "Step Into" when "importPath→OpenAPI\Client\Model.EnumTest" appears.

and then see what happens with outerEnumInteger

Enum_Test:
type: objectrequired:
- enum_string_requiredproperties:
enum_string:
type: stringenum:
- UPPER
- lower
- ''enum_string_required:
type: stringenum:
- UPPER
- lower
- ''enum_integer:
type: integerformat: int32enum:
- 1
- -1enum_number:
type: numberformat: doubleenum:
- 1.1
- -1.2outerEnum:
$ref: '#/components/schemas/OuterEnum'outerEnumInteger:
$ref: '#/components/schemas/OuterEnumInteger'outerEnumDefaultValue:
$ref: '#/components/schemas/OuterEnumDefaultValue'outerEnumIntegerDefaultValue:
$ref: '#/components/schemas/OuterEnumIntegerDefaultValue'

d07efac leads to

at some point.

But in the end

class OuterEnumInteger
{
/** * Possible values of this enum */const0 = 0;
const1 = 1;
const2 = 2;
/** * Gets allowable values of the enum * @return string[] */publicstaticfunctiongetAllowableEnumValues()
{
return [
self::0,
self::1,
self::2,
];
}
}

is generated.

Probably because after OuterEnumInteger is processed as part of EnumTest it is also processed later separately:

.

[githubERIK]

The final approach: change postProcessModelsEnum in DefaultCodegen.java to prevent adding enum variable names that are numbers. If number is detected, add a prefix NUMBER_ (see 46b9927).

Generated all samples by executing shell scripts in bin and bin/openapi3 folders using a script generate_all_sample_clients.sh.

[ybelenko]

Is this PR is still relevant or it should be closed?

[asabramo]

Hi guys,
I see this issue was reported 2 years ago.
When I use the latest:
swagger-codegen-cli-3.0.20.jar generate -l php -i api.yaml -o generated_api

I'm still seeing this issue, which means I can't use the files as they are generated - I have to add another step manually, which beats the purpose.

Oddly enough, when I use the version that the Jetbrains plugin retrieves automatically (https://repo1.maven.org/maven2/io/swagger/codegen/v3/swagger-codegen-cli/3.0.3/swagger-codegen-cli-3.0.3.jar) , the enum problem doesn't exist (but then there's another bug - BigDecimal instead of float).

Is there a version where both issues are fixed?
Is there any workaround for the enum issue?
Thanks,
Assi

[ybelenko]

@asabramo It seems that you're talking about Swagger Codegen which is different project. More information you can read at: https://github.com/OpenAPITools/openapi-generator/blob/master/docs/qna.md

About bug itself, I don't know whether it fixed or not. Maybe we should add part of the spec from related issue to samples and check PHP syntax of generated models with phplint package during CI.

[githubERIK]

@jebentier, @dkarlovi, @mandrean, @jfastnacht, @ackintosh, @ybelenko, @renepardon

Fix#1475

[githubERIK]

How is it possible that some enum 1 is transformed to ENUM_INTEGER_1 while some numeric enums just stay the same (this pull request wants (EDIT: wanted) to add the prefix "N" to these cases)?

One potential idea: enums that are pure schema objects should get the name of the schema as their prefix.

[githubERIK]

Reverted add-N-prefix changes with aa36b80 and d07efac.

The new approach, however, does not affect the clients (the generated PHP clients still have const 0=0;) which means that the changed method toEnumVarName is not called properly.

[githubERIK]

It seems to be a good idea to debug the generated jar and start with a breakpoint in DefaultCodegen and then perform "Step Into" when "importPath→OpenAPI\Client\Model.EnumTest" appears.

and then see what happens with outerEnumInteger

Enum_Test:
type: objectrequired:
- enum_string_requiredproperties:
enum_string:
type: stringenum:
- UPPER
- lower
- ''enum_string_required:
type: stringenum:
- UPPER
- lower
- ''enum_integer:
type: integerformat: int32enum:
- 1
- -1enum_number:
type: numberformat: doubleenum:
- 1.1
- -1.2outerEnum:
$ref: '#/components/schemas/OuterEnum'outerEnumInteger:
$ref: '#/components/schemas/OuterEnumInteger'outerEnumDefaultValue:
$ref: '#/components/schemas/OuterEnumDefaultValue'outerEnumIntegerDefaultValue:
$ref: '#/components/schemas/OuterEnumIntegerDefaultValue'

d07efac leads to

at some point.

But in the end

class OuterEnumInteger
{
/** * Possible values of this enum */const0 = 0;
const1 = 1;
const2 = 2;
/** * Gets allowable values of the enum * @return string[] */publicstaticfunctiongetAllowableEnumValues()
{
return [
self::0,
self::1,
self::2,
];
}
}

is generated.

Probably because after OuterEnumInteger is processed as part of EnumTest it is also processed later separately:

.

[githubERIK]

The final approach: change postProcessModelsEnum in DefaultCodegen.java to prevent adding enum variable names that are numbers. If number is detected, add a prefix NUMBER_ (see 46b9927).

Generated all samples by executing shell scripts in bin and bin/openapi3 folders using a script generate_all_sample_clients.sh.

[ybelenko]

Is this PR is still relevant or it should be closed?

[asabramo]

Hi guys,
I see this issue was reported 2 years ago.
When I use the latest:
swagger-codegen-cli-3.0.20.jar generate -l php -i api.yaml -o generated_api

I'm still seeing this issue, which means I can't use the files as they are generated - I have to add another step manually, which beats the purpose.

Oddly enough, when I use the version that the Jetbrains plugin retrieves automatically (https://repo1.maven.org/maven2/io/swagger/codegen/v3/swagger-codegen-cli/3.0.3/swagger-codegen-cli-3.0.3.jar) , the enum problem doesn't exist (but then there's another bug - BigDecimal instead of float).

Is there a version where both issues are fixed?
Is there any workaround for the enum issue?
Thanks,
Assi

[ybelenko]

@asabramo It seems that you're talking about Swagger Codegen which is different project. More information you can read at: https://github.com/OpenAPITools/openapi-generator/blob/master/docs/qna.md

About bug itself, I don't know whether it fixed or not. Maybe we should add part of the spec from related issue to samples and check PHP syntax of generated models with phplint package during CI.

[githubERIK]

@jebentier, @dkarlovi, @mandrean, @jfastnacht, @ackintosh, @ybelenko, @renepardon

Fix#1475

[githubERIK]

How is it possible that some enum 1 is transformed to ENUM_INTEGER_1 while some numeric enums just stay the same (this pull request wants (EDIT: wanted) to add the prefix "N" to these cases)?

One potential idea: enums that are pure schema objects should get the name of the schema as their prefix.

[githubERIK]

Reverted add-N-prefix changes with aa36b80 and d07efac.

The new approach, however, does not affect the clients (the generated PHP clients still have const 0=0;) which means that the changed method toEnumVarName is not called properly.

[githubERIK]

It seems to be a good idea to debug the generated jar and start with a breakpoint in DefaultCodegen and then perform "Step Into" when "importPath→OpenAPI\Client\Model.EnumTest" appears.

and then see what happens with outerEnumInteger

Enum_Test:
type: objectrequired:
- enum_string_requiredproperties:
enum_string:
type: stringenum:
- UPPER
- lower
- ''enum_string_required:
type: stringenum:
- UPPER
- lower
- ''enum_integer:
type: integerformat: int32enum:
- 1
- -1enum_number:
type: numberformat: doubleenum:
- 1.1
- -1.2outerEnum:
$ref: '#/components/schemas/OuterEnum'outerEnumInteger:
$ref: '#/components/schemas/OuterEnumInteger'outerEnumDefaultValue:
$ref: '#/components/schemas/OuterEnumDefaultValue'outerEnumIntegerDefaultValue:
$ref: '#/components/schemas/OuterEnumIntegerDefaultValue'

d07efac leads to

at some point.

But in the end

class OuterEnumInteger
{
/** * Possible values of this enum */const0 = 0;
const1 = 1;
const2 = 2;
/** * Gets allowable values of the enum * @return string[] */publicstaticfunctiongetAllowableEnumValues()
{
return [
self::0,
self::1,
self::2,
];
}
}

is generated.

Probably because after OuterEnumInteger is processed as part of EnumTest it is also processed later separately:

.

[githubERIK]

The final approach: change postProcessModelsEnum in DefaultCodegen.java to prevent adding enum variable names that are numbers. If number is detected, add a prefix NUMBER_ (see 46b9927).

Generated all samples by executing shell scripts in bin and bin/openapi3 folders using a script generate_all_sample_clients.sh.

[ybelenko]

Is this PR is still relevant or it should be closed?

[asabramo]

Hi guys,
I see this issue was reported 2 years ago.
When I use the latest:
swagger-codegen-cli-3.0.20.jar generate -l php -i api.yaml -o generated_api

I'm still seeing this issue, which means I can't use the files as they are generated - I have to add another step manually, which beats the purpose.

Oddly enough, when I use the version that the Jetbrains plugin retrieves automatically (https://repo1.maven.org/maven2/io/swagger/codegen/v3/swagger-codegen-cli/3.0.3/swagger-codegen-cli-3.0.3.jar) , the enum problem doesn't exist (but then there's another bug - BigDecimal instead of float).

Is there a version where both issues are fixed?
Is there any workaround for the enum issue?
Thanks,
Assi

[ybelenko]

@asabramo It seems that you're talking about Swagger Codegen which is different project. More information you can read at: https://github.com/OpenAPITools/openapi-generator/blob/master/docs/qna.md

About bug itself, I don't know whether it fixed or not. Maybe we should add part of the spec from related issue to samples and check PHP syntax of generated models with phplint package during CI.

[githubERIK]

@jebentier, @dkarlovi, @mandrean, @jfastnacht, @ackintosh, @ybelenko, @renepardon

Fix#1475

[githubERIK]

How is it possible that some enum 1 is transformed to ENUM_INTEGER_1 while some numeric enums just stay the same (this pull request wants (EDIT: wanted) to add the prefix "N" to these cases)?

One potential idea: enums that are pure schema objects should get the name of the schema as their prefix.

[githubERIK]

Reverted add-N-prefix changes with aa36b80 and d07efac.

The new approach, however, does not affect the clients (the generated PHP clients still have const 0=0;) which means that the changed method toEnumVarName is not called properly.

[githubERIK]

It seems to be a good idea to debug the generated jar and start with a breakpoint in DefaultCodegen and then perform "Step Into" when "importPath→OpenAPI\Client\Model.EnumTest" appears.

and then see what happens with outerEnumInteger

Enum_Test:
type: objectrequired:
- enum_string_requiredproperties:
enum_string:
type: stringenum:
- UPPER
- lower
- ''enum_string_required:
type: stringenum:
- UPPER
- lower
- ''enum_integer:
type: integerformat: int32enum:
- 1
- -1enum_number:
type: numberformat: doubleenum:
- 1.1
- -1.2outerEnum:
$ref: '#/components/schemas/OuterEnum'outerEnumInteger:
$ref: '#/components/schemas/OuterEnumInteger'outerEnumDefaultValue:
$ref: '#/components/schemas/OuterEnumDefaultValue'outerEnumIntegerDefaultValue:
$ref: '#/components/schemas/OuterEnumIntegerDefaultValue'

d07efac leads to

at some point.

But in the end

class OuterEnumInteger
{
/** * Possible values of this enum */const0 = 0;
const1 = 1;
const2 = 2;
/** * Gets allowable values of the enum * @return string[] */publicstaticfunctiongetAllowableEnumValues()
{
return [
self::0,
self::1,
self::2,
];
}
}

is generated.

Probably because after OuterEnumInteger is processed as part of EnumTest it is also processed later separately:

.

[githubERIK]

The final approach: change postProcessModelsEnum in DefaultCodegen.java to prevent adding enum variable names that are numbers. If number is detected, add a prefix NUMBER_ (see 46b9927).

Generated all samples by executing shell scripts in bin and bin/openapi3 folders using a script generate_all_sample_clients.sh.

[ybelenko]

Is this PR is still relevant or it should be closed?

[asabramo]

Hi guys,
I see this issue was reported 2 years ago.
When I use the latest:
swagger-codegen-cli-3.0.20.jar generate -l php -i api.yaml -o generated_api

I'm still seeing this issue, which means I can't use the files as they are generated - I have to add another step manually, which beats the purpose.

Oddly enough, when I use the version that the Jetbrains plugin retrieves automatically (https://repo1.maven.org/maven2/io/swagger/codegen/v3/swagger-codegen-cli/3.0.3/swagger-codegen-cli-3.0.3.jar) , the enum problem doesn't exist (but then there's another bug - BigDecimal instead of float).

Is there a version where both issues are fixed?
Is there any workaround for the enum issue?
Thanks,
Assi

[ybelenko]

@asabramo It seems that you're talking about Swagger Codegen which is different project. More information you can read at: https://github.com/OpenAPITools/openapi-generator/blob/master/docs/qna.md

About bug itself, I don't know whether it fixed or not. Maybe we should add part of the spec from related issue to samples and check PHP syntax of generated models with phplint package during CI.

[githubERIK]

@jebentier, @dkarlovi, @mandrean, @jfastnacht, @ackintosh, @ybelenko, @renepardon

Fix#1475

[githubERIK]

How is it possible that some enum 1 is transformed to ENUM_INTEGER_1 while some numeric enums just stay the same (this pull request wants (EDIT: wanted) to add the prefix "N" to these cases)?

One potential idea: enums that are pure schema objects should get the name of the schema as their prefix.

[githubERIK]

Reverted add-N-prefix changes with aa36b80 and d07efac.

The new approach, however, does not affect the clients (the generated PHP clients still have const 0=0;) which means that the changed method toEnumVarName is not called properly.

[githubERIK]

It seems to be a good idea to debug the generated jar and start with a breakpoint in DefaultCodegen and then perform "Step Into" when "importPath→OpenAPI\Client\Model.EnumTest" appears.

and then see what happens with outerEnumInteger

Enum_Test:
type: objectrequired:
- enum_string_requiredproperties:
enum_string:
type: stringenum:
- UPPER
- lower
- ''enum_string_required:
type: stringenum:
- UPPER
- lower
- ''enum_integer:
type: integerformat: int32enum:
- 1
- -1enum_number:
type: numberformat: doubleenum:
- 1.1
- -1.2outerEnum:
$ref: '#/components/schemas/OuterEnum'outerEnumInteger:
$ref: '#/components/schemas/OuterEnumInteger'outerEnumDefaultValue:
$ref: '#/components/schemas/OuterEnumDefaultValue'outerEnumIntegerDefaultValue:
$ref: '#/components/schemas/OuterEnumIntegerDefaultValue'

d07efac leads to

at some point.

But in the end

class OuterEnumInteger
{
/** * Possible values of this enum */const0 = 0;
const1 = 1;
const2 = 2;
/** * Gets allowable values of the enum * @return string[] */publicstaticfunctiongetAllowableEnumValues()
{
return [
self::0,
self::1,
self::2,
];
}
}

is generated.

Probably because after OuterEnumInteger is processed as part of EnumTest it is also processed later separately:

.

[githubERIK]

The final approach: change postProcessModelsEnum in DefaultCodegen.java to prevent adding enum variable names that are numbers. If number is detected, add a prefix NUMBER_ (see 46b9927).

Generated all samples by executing shell scripts in bin and bin/openapi3 folders using a script generate_all_sample_clients.sh.

[ybelenko]

Is this PR is still relevant or it should be closed?

[asabramo]

Hi guys,
I see this issue was reported 2 years ago.
When I use the latest:
swagger-codegen-cli-3.0.20.jar generate -l php -i api.yaml -o generated_api

I'm still seeing this issue, which means I can't use the files as they are generated - I have to add another step manually, which beats the purpose.

Oddly enough, when I use the version that the Jetbrains plugin retrieves automatically (https://repo1.maven.org/maven2/io/swagger/codegen/v3/swagger-codegen-cli/3.0.3/swagger-codegen-cli-3.0.3.jar) , the enum problem doesn't exist (but then there's another bug - BigDecimal instead of float).

Is there a version where both issues are fixed?
Is there any workaround for the enum issue?
Thanks,
Assi

[ybelenko]

@asabramo It seems that you're talking about Swagger Codegen which is different project. More information you can read at: https://github.com/OpenAPITools/openapi-generator/blob/master/docs/qna.md

About bug itself, I don't know whether it fixed or not. Maybe we should add part of the spec from related issue to samples and check PHP syntax of generated models with phplint package during CI.