Special characters are not escaped in generated code

The following schema is valid, and works in a compliant draft 07 implementation

{
  "$schema": "http://json-schema.org/draft-07/schema#",
  "$id": "x",
  "definitions": {
    "(+[{*#$%^&~`';:,.<> \"?!\\n}])": {
      "definitions": {
        "array": { "type": "array" },
        "object": { "type": "object" }
      },
      "anyOf": [
        { "$ref": "#/definitions/(+[{*#$%^&~`';:,.<> \"?!\\n}])/definitions/array" },
        { "$ref": "#/definitions/(+[{*#$%^&~`';:,.<> \"?!\\n}])/definitions/object" }
      ]
    }
  },
  "properties": {
    "x": { "$ref": "#/definitions/(+[{*#$%^&~`';:,.<> \"?!\\n}])" }
  }
}

import fastjsonschema, json
x = fastjsonschema.compile(json.load(open("./nested.schema.json", "r")))
Traceback (most recent call last):
  File "<stdin>", line 1, in <module>
  File "/lib/python3.7/site-packages/fastjsonschema/__init__.py", line 103, in validate
    return compile(definition, handlers, formats)(data)
  File "/lib/python3.7/site-packages/fastjsonschema/__init__.py", line 167, in compile
    exec(code_generator.func_code, global_state)
  File "<string>", line 10
    validate_x__definitions_(+[{*_$_^&~`';_,_<> ?!\n}])(data__x)

Some characters are properly removed, including the \" and #, but many are not allowed in a Python name.

Simply replacing them with _ is not such a good solution because then a property named # is the same as % and _ etc.

Using things like _open_paren to escape ( etc, is somewhat ok, until a user makes a property named _open_paren, and then the names ( and _open_paren are wrongly considered the same.

A solution I have seen before (in Brat code gen) is to give all special characters unique names, and then to make user-inputted names like identifiers surrounded by __ 2 underscores, which are translated into _dunder_ (for double-underscore), and then ensure that user input is expanded in a way that can never overlap with a different input.

I cannot find where in the JSON Schema specification it says that any character other than / is allowed as an identifier / key name, but I am sure they are. NUL bytes are also allowed except I don’t know of any Python JSON parsers that handle them properly, so I don’t really care.

Also, double-escaping the characters doesn’t help (one backslash alone is invalid to the JSON parser):

{
  "$schema": "http://json-schema.org/draft-07/schema#",
  "$id": "x",
  "definitions": {
    "\\(p\\)": {}
  },
  "properties": {
    "x": { "$ref": "#/definitions/\\(p\\)" }
  }
}

x = fastjsonschema.compile(json.load(open("./nested.schema.json", "r")))
Traceback (most recent call last):
  File "<stdin>", line 1, in <module>
  File "/home/cat/Sync/projects/git/pyrat/.venv/lib/python3.7/site-packages/fastjsonschema/__init__.py", line 167, in compile
    exec(code_generator.func_code, global_state)
  File "<string>", line 10
    validate_x__definitions_\(p\)(data__x)
                                         ^
SyntaxError: unexpected character after line continuation character

[horejsek]

Thanks for the report. I have an idea how to solve it–by making hash of the original name.

[horejsek]

For now I fixed it with simple removing of invalid characters in v2.14.2 which is also used for long time for property names. Let me know if this is not enough for now.