Endpoint throws TypeError of JSON encoding, using mutation with input types.

[EdgarDegas]

🐜 Bug Report

I'm using code generation. And I send mutations like this:

op = mutation_log_ticket_check()
endpoint(op, variables={
    "ticketId": ticket_id,
    "act": act,
    "reason": reason,
    "info": info,
})

This throws an error TypeError: Object of type TicketCheckLogInfo is not JSON serializable, where info is an instance of an input type TicketCheckLogInfo:

class TicketCheckLogInfo(sgqlc.types.Input):
    __schema__ = types
    __field_names__ = ('summary', 'confidence', 'comment', 'is_tech', 'category')
    summary = sgqlc.types.Field(String, graphql_name='summary')
    confidence = sgqlc.types.Field(Int, graphql_name='confidence')
    comment = sgqlc.types.Field(String, graphql_name='comment')
    is_tech = sgqlc.types.Field(Boolean, graphql_name='isTech')
    category = sgqlc.types.Field(String, graphql_name='category')

Expected Behavior

I'm using code generation, so I suppose instances of generated input types would fit in the variable slots of operations, which are code that's also generated.

Please tell me if I'm using this lib in a wrong way and, if so, how to integrate input types into operations correctly.

Right now I'm using this as a workaround:

"info": TicketCheckLogInfo.__to_json_value__(info),

Current Behavior

Apparently, HTTPEndpoint.__call__ just dumps the variables dict into JSON.