The generated output should (optionally?) be a complete file ready for import into other projects, and include imports like:

from pydantic import BaseModel
from types import List, ...

...

Is your feature request related to a problem? Please describe.
Didn't use this library yet, just looked at code and saw that BaseModel is hardcoded in templates as base class for models:

Describe the solution you'd like
Add ability to choose the name of base class of models (and even create one automatically)
This would be helpful if I wan't to have same config, methods, etc for all subclasses

Describe alternatives you've considered
This can be approached by renaming BaseModel in a result manually, but if it not hard to implement I'd preffered to have this built-in

Additional context
Add any other context or screenshots about the feature request here.

I will try to publish the package soon.

Is your feature request related to a problem? Please describe.
OpenAPI or JsonSchema may describe key of objects as camelCase.
But, attributes of python should be snake_case.
The generator should provide a feature that converts filed names CamelCase to sanke_case.

Describe the solution you'd like
I would the feature give the generator a command-line option

This PR supports byte of string format.

OpenAPI expects that byte is base64-encoded characters.
https://swagger.io/docs/specification/data-models/data-types/#string

But, I don't know how do we treat encoded/decoded values like base64 in Pydantic.
I opened this issue. However, the problem is not resolved yet.
pydantic/pydantic#692

Related issues

#30

Format types

• date
• date-time
• password
• byte
• binary
• email
• uuid
• uri
• hostname
• ipv4
• ipv6
• deciaml

It would be great to support python's decimal.Decimal type.

Describe the solution you'd like
There's some examples, how it could be implemented on the OpenAPI side:

{"type": "string", "format": "decimal"}
{"type": "number", "format": "decimal"}

Describe the bug
Default value is double quoted if string

To Reproduce

Example schema:

{
    "definitions": {
        "io.k8s.api.core.v1.ComponentStatus": {
            "description": "ComponentStatus (and ComponentStatusList) holds the cluster validation info.",
            "properties": {
                "apiVersion": {
                    "description": "APIVersion defines the versioned schema of this representation of an object. Servers should convert recognized schemas to the latest internal value, and may reject unrecognized values. More info: https://git.k8s.io/community/contributors/devel/sig-architecture/api-conventions.md#resources",
                    "type": "string",
                    "default": "v1"
                },
                "kind": {
                    "description": "Kind is a string value representing the REST resource this object represents. Servers may infer this from the endpoint the client submits requests to. Cannot be updated. In CamelCase. More info: https://git.k8s.io/community/contributors/devel/sig-architecture/api-conventions.md#types-kinds",
                    "type": "string",
                    "enum": [
                        "ComponentStatus"
                    ],
                    "default": "ComponentStatus"
                }
            },
            "type": "object",
            "x-kubernetes-group-version-kind": [
                {
                    "group": "",
                    "kind": "ComponentStatus",
                    "version": "v1"
                }
            ]
        }
    }
}

Used commandline:

$ datamodel-codegen --input k8s_1.16.4_patched.json --input-file-type jsonschema --output test_

Expected behavior

class ComponentStatus(BaseModel):
    apiVersion: Optional[str] = Field(
        "v1",
        description="...",
    )

But got:

class ComponentStatus(BaseModel):
    apiVersion: Optional[str] = Field(
        "'v1'",
        description="...",
    )

Version:

• OS: Linux
• Python Verison: 3.8
• datamodel-code-generator git master, 2cc4d0d

Additional context
This is a regression, was working a few commits ago.

Describe the bug

To Reproduce

Example schema:
example file https://raw.githubusercontent.com/kubernetes/kubernetes/1.16.4/api/openapi-spec/swagger.json

Generated code in io/k8s/apimachinery/pkg/version.py:

from __future__ import annotations

from pydantic import BaseModel


class Info(BaseModel):
    build_date: str = Field(..., alias="buildDate")
    compiler: str
    git_commit: str = Field(..., alias="gitCommit")
    git_tree_state: str = Field(..., alias="gitTreeState")
    git_version: str = Field(..., alias="gitVersion")
    go_version: str = Field(..., alias="goVersion")
    major: str
    minor: str
    platform: str

Expected behavior

from pydantic import BaseModel, Field

Version:

• OS: Linux
• Python Verison: 3.8
• datamodel-code-generator master

Is your feature request related to a problem? Please describe.

Remote $ref are not resolved

Describe the solution you'd like

Remote $ref to be resolved

Describe alternatives you've considered

Remote $refs are downloaded and converted to local refs by an external component

Additional context

components:
  schemas:
    Problem:
      $ref: "https://opensource.zalando.com/problem/schema.yaml#/Problem"

TODO

Run a command to check type

• mypy

The generated model does not have the description from source schema. It is useful to have this description in code (for example to provide help text).

There is no document of the custom template.
We should write a document on how to use the feature.

Related issues

#144

TODO

Run a command to check code

• flake8 or black

Describe the bug
I think it is allowed to use oneOf inside properties but it won't work.

To Reproduce
Steps to reproduce the behavior:

small.json

{
    "definitions": {
      "io.k8s.api.admissionregistration.v1.MutatingWebhook": {
        "properties": {
          "timeout": {
            "oneOf": [
                {
                  "type": "string"
                },
                {
                  "type": "integer"
                }
            ]
          }
        },  
        "type": "object"
      }
    }
}

$ datamodel-codegen --input small.json --input-file-type jsonschema 
Traceback (most recent call last):
  [...]
  File "datamodel_code_generator/parser/jsonschema.py", line 158, in get_data_type
    raise ValueError(f"invalid schema object {obj}")
ValueError: invalid schema object items=None uniqueItem=None type=None format=None pattern=None minLength=None maxLength=None minimum=None maximum=None multipleOf=None exclusiveMaximum=None exclusiveMinimum=None additionalProperties=None anyOf=[] allOf=[] enum=[] writeOnly=None properties=None required=[] ref=None nullable=False x_enum_varnames=[] description=None title=None example=None examples=None default=None

Its me again and I have a new crazy JSON here.
Sometimes I have JSON which contains a key starting with a number.

To Reproduce
Example JSON:

{
    "Values": {
        "1 Step": "42",
        "2 Step": "23"
    }
}

Used commandline:

$ datamodel-codegen --input jsons/numbers.json --input-file-type json --output models/numbers.py --target-python-version 3.6

Expected behavior
Model is created. For the keys starting with a number there should be a underscore "_" added at the beginning of the string.

Current behavior
An error is thrown:

Traceback (most recent call last):
  File "/home/flz/.local/share/virtualenvs/modelCreate-nvqMaMoH/lib/python3.6/site-packages/datamodel_code_generator/__main__.py", line 122, in main
    validation=namespace.validation,
  File "/home/flz/.local/share/virtualenvs/modelCreate-nvqMaMoH/lib/python3.6/site-packages/datamodel_code_generator/__init__.py", line 179, in generate
    result = parser.parse()
  File "/home/flz/.local/share/virtualenvs/modelCreate-nvqMaMoH/lib/python3.6/site-packages/datamodel_code_generator/parser/base.py", line 317, in parse
    body = format_code(body, self.target_python_version)
  File "/home/flz/.local/share/virtualenvs/modelCreate-nvqMaMoH/lib/python3.6/site-packages/datamodel_code_generator/format.py", line 26, in format_code
    code = apply_black(code, python_version)
  File "/home/flz/.local/share/virtualenvs/modelCreate-nvqMaMoH/lib/python3.6/site-packages/datamodel_code_generator/format.py", line 46, in apply_black
    string_normalization=not config.get("skip-string-normalization", True),
  File "/home/flz/.local/share/virtualenvs/modelCreate-nvqMaMoH/lib/python3.6/site-packages/black.py", line 725, in format_str
    src_node = lib2to3_parse(src_contents.lstrip(), mode.target_versions)
  File "/home/flz/.local/share/virtualenvs/modelCreate-nvqMaMoH/lib/python3.6/site-packages/black.py", line 836, in lib2to3_parse
    raise exc from None
black.InvalidInput: Cannot parse: 5:5:     1_Step: str = Field(..., alias='1 Step')

Version:

• OS: Ubuntu 20.04
• Python Verison: 3.6.11
• datamodel-code-generator Version 0.5.13

Additional context
I think in Python variable names must not start with a number.

The code-generator creates models for old-style pydantic (under v1.0) eg: UrlStr
I want to change the supporting version of pydantic to v1.0 or later.

Changes

• UrlStr to AnyStr for URI
• use Field()

I considered we should drop the old-pydantic version.
If someone has any idea then, please tell me it.

Related PRs

#100
#99

Describe the bug
Parsing yaml OAS3 files does not work

To Reproduce
See #183 .

Example schema:

components:
  schemas:
    Problem:
      $ref: https://teamdigitale.github.io/openapi/0.0.6/definitions.yaml#/schemas/Problem

Used commandline:

$ datamodel-codegen test.yaml
``

**Expected behavior**
A generated Problem class

**Version:**
 - OS: linux
 - Python Verison: 3.7
 - datamodel-code-generator Version latest

Related PR

#36

The code generator should handle json schema.
The way is to implement JSON Shema Parser.

In our project we process a large number of JSON files from external source. We want to use pydantic for data models and also to check the structure of incoming files. The JSON files has spaces and special characters in the keys.

Example:

{
    "Serial Number": "A12345678",
    "Timestamp": "2020-05-26T12:15:25.792741Z",
    "Data": {
        "Length (m)": 12.34,
        "Symmetric deviation (%)": 12.216564148290807,
        "Total running time (s)": 974,
        "Mass (kg)": 42.23,
        "Initial parameters": {
            "V1": 123,
            "V2": 456
        }
    }
}

The keys with space and brackets works perfect. They are generated as Length__m_: float = Field(..., alias='Length (m)') for example.
The generator don't likes the "Data.Initial parameters" key in the example above. It is the name of the sub structure and would result in a class name with space The following error is raised:
black.InvalidInput: Cannot parse: 6:14: class Initial parameters(BaseModel):

It would be cool, if the generator would support such files. Is this possible to add this feature?

We need unittest and cover all.

When trying to convert file data.json to model.py, I get a
keyerror on line 171 of jsonschema.py when the type in types is 'array'.

To Reproduce

Example schema:

I really like the idea behind this -- jinja is a lot easier to work with than openapi-generator templates and the java-based generators.

Would you have any interest in incorporating api-client-generation capability into this codebase? I would be happy to put a little effort into transforming the API client and templated API generation from https://github.com/dmontagu/fastapi_client into something that would combine better with this code base. That way, it would 1) have fewer generation-time dependencies 2) an easier-to-work-with templating language, and 3) leverage the work you've already put in for model generation.

But I would also understand if you'd rather keep this repository more limited in scope and targeted specifically around model generation.

I am generating models from JSON Data which comes from an external source. In my JSON file there is a element which contains a other sub element with the same name.

Example JSON

{
    "recursive": {
        "sub": {
            "recursive": {
                "value": 42.23
            }
        }
    }
}

Used commandline:

$ datamodel-codegen --input jsons/recursive.json --input-file-type json --output models/recursive.py --target-python-version 3.6

Expected behavior
The Model should look like this.

class Recursive(BaseModel):
    sub: 'Sub'

class Recursive2(BaseModel):
    sub: float

class Sub(BaseModel):
    recursive: 'Recursive2'

class Model(BaseModel):
    recursive: 'Recursive'

But it looks like this and has an indirect recursion in Recursive->Sub->Recursive... :

class Recursive(BaseModel):
    sub: 'Sub'

class Sub(BaseModel):
    recursive: 'Recursive'

class Model(BaseModel):
    recursive: 'Recursive'

Version:

• OS: Ubuntu 20.04
• Python Verison: 3.6
• datamodel-code-generator Version 0.5.13

Describe the bug
I am not yet sure how to reproduce which state, as I have two different observation:

Example 1:

Generated code:

from ...apimachinery.pkg.apis.meta import v1
from ..core import v1

(from: https://github.com/FlorianLudwig/pdk8s/blob/e9b5903c4f8cf584fe60a4a84709a42f7e63b771/pdk8s/gen/io/k8s/api/networking/v1beta1.py#L12)

Example 2:

Generated code:

from ...apimachinery.pkg.apis.meta import v1
from ..core import v1 as v1_1

(from: https://github.com/FlorianLudwig/pdk8s/blob/e9b5903c4f8cf584fe60a4a84709a42f7e63b771/pdk8s/gen/io/k8s/api/extensions/v1beta1.py)

In this example, the second import is renamed to avoid the collision but the references are not. Resulting in code trying to access module members that are in v1_1 but refer to v1. (Example)

Expected behavior

The second example is a step in the right direction (or maybe import more of the package name, in this example meta.v1 instead of just v1). But the references must be updated.

Version:

• datamodel-code-generator: git master

The code generator supports only one file by $ref
$ref should reference external files.

Motivation

#28
@prostomarkeloff said

Firstly, i think about $ref fields, without it, it is easy. I checked readme and saw information what $ref supported only in 1 file, we should to "fix" that

Documents

https://swagger.io/docs/specification/using-ref/

Describe the bug
Hi! It seems that when using an array of enum (string type), the code is not being generated properly. Instead of making a List of enums, it always allows any string because it makes a List[str].

To Reproduce

Example schema:

openapi: 3.0.0
info:
  title: datamodel-code-generator bug example
components:
  schemas:
    Type1:
      type: array
      items:
        type: string
        enum:
          - enumOne
          - enumTwo

    Type2:
      type: string
      enum:
        - enumFour
        - enumFive

Code generated:

from __future__ import annotations

from enum import Enum
from typing import List

from pydantic import BaseModel


class Type1(BaseModel):
    __root__: List[str]


class Type2(Enum):
    enumFour = 'enumFour'
    enumFive = 'enumFive'

Used commandline:

$ datamodel-codegen --input input.yaml --output tmp.py

Expected behavior
Expected behaviour would be that inside of Type1, there should be List[Enum] where Enum is a class that has the correct enum types allowed.

Version:

• OS: MacOS 10.15
• Python Verison: 3.7.7
• datamodel-code-generator Version: 0.5.4

Or is there an alternative way of achieving the above that I am missing?
The goal is to essentially have a class that accepts a List of enum.
Thanks!

Describe the bug
I want to create pydantic models for the Spotify api and tried to parse an example response (see here).

• datamode-codegen 0.4.4
• python 3.8

To Reproduce

Download the gist, install datamodel-codegen and run the following command.

The command I am using to generate the the models is

datamodel-codegen --input-file-type jsonschema --input api_response.json --out output.py

Screenshots

I've tried to run this against Kubernetes OpenAPI Spec but ran into some errors.

curl -O https://raw.githubusercontent.com/kubernetes/kubernetes/master/api/openapi-spec/swagger.json
datamodel-codegen --input swagger.json --input-file-type=openapi --output=mode.py

[... error abridged ...]
  File "/proj/.venv/lib/python3.7/site-packages/datamodel_code_generator/parser/base.py", line 164, in parse
    self.parse_raw()
  File "/proj/.venv/lib/python3.7/site-packages/datamodel_code_generator/parser/openapi.py", line 12, in parse_raw
    spec_string=self.text, backend='openapi-spec-validator'
  File "/proj/.venv/lib/python3.7/site-packages/prance/__init__.py", line 106, in __init__
    self.parse()
  File "/proj/.venv/lib/python3.7/site-packages/prance/__init__.py", line 137, in parse
    self._validate()
  File "/proj/.venv/lib/python3.7/site-packages/prance/__init__.py", line 172, in _validate
    validator(parsed)
  File "/proj/.venv/lib/python3.7/site-packages/prance/__init__.py", line 237, in _validate_openapi_spec_validator
    raise_from(ValidationError, type_ex)
  File "/proj/.venv/lib/python3.7/site-packages/prance/util/exceptions.py", line 16, in raise_from
    raise klass(*from_value.args) from from_value
prance.ValidationError: unhashable type: 'dict'

I also tried the JSON Schema version but hit different errors.

pip install openapi2jsonschema
openapi2jsonschema swagger.json --kubernetes
cd schemas
datamodel-codegen --input pod.yaml --input-file-type=jsonschema --output=mode.py

Traceback (most recent call last):
  File "/proj/.venv/bin/datamodel-codegen", line 8, in <module>
    sys.exit(main())
  File "/proj/.venv/lib/python3.7/site-packages/datamodel_code_generator/__main__.py", line 168, in main
    result = parser.parse()
  File "/proj/.venv/lib/python3.7/site-packages/datamodel_code_generator/parser/base.py", line 164, in parse
    self.parse_raw()
  File "/proj/.venv/lib/python3.7/site-packages/datamodel_code_generator/parser/jsonschema.py", line 472, in parse_raw
    self.parse_raw_obj(obj_name, raw_obj)
  File "/proj/.venv/lib/python3.7/site-packages/datamodel_code_generator/parser/jsonschema.py", line 455, in parse_raw_obj
    obj = JsonSchemaObject.parse_obj(raw)
  File "/proj/.venv/lib/python3.7/site-packages/pydantic/main.py", line 402, in parse_obj
    return cls(**obj)
  File "/proj/.venv/lib/python3.7/site-packages/pydantic/main.py", line 283, in __init__
    raise validation_error
pydantic.error_wrappers.ValidationError: 2 validation errors for JsonSchemaObject
properties -> apiVersion -> type
  str type expected (type=type_error.str)
properties -> kind -> type
  str type expected (type=type_error.str)

Can this library support Kubernetes Open API Spec or JSON Schema?

The OpenAPI v3.0.1 schema allows JsonSchemaObject's additionalProperties to be a boolean or a JsonSchemaObject.

Specifically, the spec (https://github.com/OAI/OpenAPI-Specification/blob/master/versions/3.0.1.md#parameterObject) says:

"additionalProperties - Value can be boolean or object. Inline or referenced schema MUST be of a Schema Object and not a standard JSON Schema."

However the definition at L51 of jsonschema.py has:
additionalProperties: Optional['JsonSchemaObject'].

This fails on schemas with additionalProperties: false.

Suggested fix is to change L51 of jsonschema.py to:
additionalProperties: Union['JsonSchemaObject', bool, None].

add documents include examples on each feature.

It would be good if we can identify that a similar model exists and use that model instead of creating a new model, especially when the FIELDS and DATA-TYPE are same.

For example, when my data is as follows,

{
    "Arm Right": {
        "Joint 1": 5,
        "Joint 2": 3, 
        "Joint 3": 66
    }, 
    "Arm Left": {
        "Joint 1": 55,
        "Joint 2": 13, 
        "Joint 3": 6
    }
}

we get the following output,

class ArmRight(BaseModel):
    Joint_1: int = Field(..., alias='Joint 1')
    Joint_2: int = Field(..., alias='Joint 2')
    Joint_3: int = Field(..., alias='Joint 3')

class ArmLeft(BaseModel):
    Joint_1: int = Field(..., alias='Joint 1')
    Joint_2: int = Field(..., alias='Joint 2')
    Joint_3: int = Field(..., alias='Joint 3')

class Model(BaseModel):
    Arm_Right: ArmRight = Field(..., alias='Arm Right')
    Arm_Left: ArmLeft = Field(..., alias='Arm Left')

Describe the solution you'd like
I think it would be good if we can detect that models for ArmLeft and ArmRight are the same, and then to just use one of these models. For example,

class ArmRight (BaseModel):
    Joint_1: int = Field(..., alias='Joint 1')
    Joint_2: int = Field(..., alias='Joint 2')
    Joint_3: int = Field(..., alias='Joint 3')

class Model(BaseModel):
    Arm_Right: ArmRight = Field(..., alias='Arm Right')
    Arm_Left: ArmRight = Field(..., alias='Arm Left')

Describe alternatives you've considered
Currently, I just delete manually. But, we have a big system with many similar structures in JSON and many time we need to manually delete about 1000 models. :(

Thank you very much for this excellent tool.

If a field is nullable (https://swagger.io/docs/specification/data-models/data-types/#null) use the Optional type hint (https://pydantic-docs.helpmanual.io/#pep-484-types)

support Enums
The generator should handle Enums types

Related Links

OpenAPI

https://swagger.io/docs/specification/data-models/enums/

Pydantic

https://pydantic-docs.helpmanual.io/#choices

p1c2u/openapi-spec-validator as great as it is, does not properly validate valid OpenAPI 3 specs (at least not all of them) - since model generation relies on validation this library is also "broken" (I tried to avoid this word but my English is not good enough)

The solution could be:

1. us taking matters in our own hands
2. using a proper validation tool like https://www.npmjs.com/package/swagger-parser. I am super uneasy with mixing JS with Python but nowadays running a tool in a pre-made container is a valid option
3. port the thing to Python
   It's easy to say but I am only capable of doing the second. What you think?

datamodel-code-generator is at this point needed by the Internet - being able to generate FastAPI definitions from OpenAPI 3 will make backend PyDev work much easier and API first is generally a good approach.

Like I said I'm not capable enough of making it on my own - maybe together will be able to figure things out. If you see this "issue" valid of course.

datamodel-codegen is gaining a lot of command-line options which is great, but it can make the invocation quite lengthy.

How about adding support for a pyproject.toml section, e.g.

[tool.datamodel-codegen]
field-constraints = true
snake-case-field = true
strip-default-none = false
target-python-version = "3.7"

I'm happy to add this if there is desire for this feature :)

Describe the bug
having a - inside a definition's name results in an invalid python module name.

Expected behavior

Replace - in names with _.

Using the JSONSchema way to declare a pattern for a string property fails on datamodel conversion.

Based on the JSONSchema docs, the right way to declare a regex to validate the property is:

{
"rasen":
   "type": "string",
   "pattern": "^(\\([0-9]{3}\\))?[0-9]{3}-[0-9]{4}$"
}

But using this in a property make datamodel-code-generator crash with a InvalidInput error because it cannot parse the arguments to use in the generated Field:

  File "/usr/lib/python3.8/site-packages/black.py", line 725, in format_str
    src_node = lib2to3_parse(src_contents.lstrip(), mode.target_versions)
  File "/usr/lib/python3.8/site-packages/black.py", line 836, in lib2to3_parse
    raise exc from None
black.InvalidInput: Cannot parse: 22:33:     rasen: Optional[constr(regex=^(\([0-9]{3}\))?[0-9]{3}-[0-9]{4}$)] = None

I tried to do the same with a yml file and found that it crash in the same point.

ransen_property:
        type: string
        description: Test
        pattern: "[\/].*$"

The traceback in this case was:

  File "/usr/lib/python3.8/site-packages/black.py", line 725, in format_str
    src_node = lib2to3_parse(src_contents.lstrip(), mode.target_versions)
  File "/usr/lib/python3.8/site-packages/black.py", line 836, in lib2to3_parse
    raise exc from None
black.InvalidInput: Cannot parse: 35:28:     __root__: constr(regex=[/].*$)

I tracked the issue back to the type validator for strings and found that striping the quotes from the regex fires this error. I will make a PR with a suggested workaround to fix this.

Describe the bug

I think the tool might generate some = None to many, for example here:

https://github.com/koxudaxi/datamodel-code-generator/blob/master/tests/test_main.py#L107

I do have a WIP fix but maybe I am wrong and it should be there?

This code generator can't generate some model in edge-case.
I'm implementing a model resolver to fix the problem.
The model resolver should cover below.

• manage collected model (include ref )
• manage unique model names
• validate valid Python name
• resolve remote or other local files
• manage the scoped model name for import statement

Sorry, I'm busy with the main job this month.
But, I implement the feature step by step. I will get the time for this project 😄

Related issues

#170
#169
#168
#156
#145

Related PR

#36

I expect to handle Minimum and Maximum include exclusiveMinimum and exclusiveMaximum , multipleOf

Related PR

#36

Problem

It is an assumption that there is only a single _ in field names. This can happen if there are two spaces in a name (https://github.com/koxudaxi/datamodel-code-generator/blob/master/datamodel_code_generator/parser/base.py#L142), or the field itself contains two __.

So I think making an assumption that anything with _ is valid snake case is incorrect.

Proposed solution

Change https://github.com/koxudaxi/datamodel-code-generator/blob/master/datamodel_code_generator/parser/base.py#L45 to

return prefix + ''.join(x[0].upper() + x[1:] for x in word.split('_') if x)

That way if there are multiple _ they're skipped and the camelcase is still correctly created.

We should change the expected data of tests to external files.

example
https://github.com/koxudaxi/fastapi-code-generator/blob/983c0bcb8978969a591bab242fbb6bcfa3c8e039/tests/test_generate.py

Related issues

#159

Describe the bug
Invalid code generated for polymorphic array in Swagger spec.

To Reproduce
Using the following Swagger YAML:

openapi: 3.0.0
info:
  title: datamodel-code-generator bug example
components:
  schemas:
    Container:
      allOf:
        - type: object
          required:
            - contents
          properties:
            contents:
              type: array
              items:
                anyOf:
                  - $ref: '#/components/schemas/Type1'
                  - $ref: '#/components/schemas/Type2'
    Type1:
      type: object
      properties:
        prop:
          type: string
    Type2:
      type: object
      properties:
        prop:
          type: string

The generated code:

# generated by datamodel-codegen:
#   filename:  swagger-borked.yaml
#   timestamp: 2020-06-10T22:51:26+00:00

from __future__ import annotations

from typing import List, Optional

from pydantic import BaseModel


class Type1(BaseModel):
    prop: Optional[str] = None


class Type2(BaseModel):
    prop: Optional[str] = None


class Container(BaseModel):
    contents: List[Type1, Type2]

The last line is invalid and causes a runtime error.

TypeError: Too many parameters for typing.List; actual 2, expected 1

Expected behavior
The generated code should be a list containing a union of the types.

class Container(BaseModel):
    contents: List[Union[Type1, Type2]]

This occurs using the latest version of datamodel-code-generator, 0.5.2.

Thanks for the tool! It's very handy. I tried to figure out how to fix this myself but found myself going in circles.

Describe the bug

custom_template_dir seems to be ignored:

datamodel_code_generator.generate(custom_template_dir="some/path")

Version:

• datamodel-code-generator Version: git master

The PR is to convert plurals to singular for array models.

Detail

code-generator have to define an item from an array.
But, the code-generator don't know the valid singular name.
I think the code-generator can resolve the singular name with this library.
https://pypi.org/project/inflect/