Modules/_json.c

Source:

cpython 3.14 @ ab2d84fe1023/Modules/_json.c ↗

Modules/_json.c provides a C acceleration for the two performance-critical parts of json: string scanning (decoder) and string encoding (encoder). Lib/json/decoder.py and Lib/json/encoder.py import scanstring and encode_basestring/encode_basestring_ascii from _json when the C extension is available.

Map

Lines	Symbol	Role
1-100	Encoder object, encode_string	JSON string output
101-400	scanstring_impl	JSON string input (scanner)
401-800	encoder_encode_dict, encoder_encode_list	Recursive object serialization
801-1100	encoder_listencode_obj	Type dispatch: int, float, str, list, dict
1101-1500	Encoder_call	__call__ for the JSONEncoder C type
1501-1800	PyModuleDef, type definitions	Module and type registration

Reading

scanstring_impl: JSON string decoder

The C decoder scans a JSON string literal, handling escape sequences (\n, \t, \uXXXX). It is significantly faster than the Python equivalent for large strings because it avoids per-character Python overhead.

// CPython: Modules/_json.c:235 scanstring_impl
static PyObject *
scanstring_impl(PyObject *pystr, Py_ssize_t end, int strict)
{
    ...
    while (1) {
        /* Fast path: scan for the next special character */
        Py_ssize_t i = next_special(buf, end, len);
        if (buf[i] == '"') {
            /* end of string */
            break;
        }
        else if (buf[i] == '\\') {
            /* escape sequence */
            char c = buf[i+1];
            switch (c) {
            case '"': ...
            case '\\': ...
            case 'n': chunks.append('\n'); ...
            case 'u': /* \uXXXX */ ...
            }
        }
    }
    ...
}

encode_basestring and encode_basestring_ascii

encode_basestring(s) produces a JSON string literal from a Python str, escaping only the required characters (", \, control chars). encode_basestring_ascii also escapes non-ASCII characters as \uXXXX.

encoder_encode_dict

Recursively encodes a Python dict as a JSON object. It iterates dict.items(), encodes each key (must be a string) and value, and joins them with ,. The sort_keys option sorts the keys before encoding.

// CPython: Modules/_json.c:650 encoder_encode_dict
static int
encoder_encode_dict(PyEncoderObject *s, _PyUnicodeWriter *writer,
                    PyObject *dct, Py_ssize_t indent_level)
{
    ...
    if (s->sort_keys) {
        items = PyMapping_Items(dct);
        PyList_Sort(items);
    } else {
        items = PyMapping_Items(dct);
    }
    ...
    for each (key, value) in items:
        encode_key(key);
        writer->write(": ");
        encoder_listencode_obj(value);
        ...
}

gopy notes

Status: json is a high-priority port. The Go standard library encoding/json is available but not Python-compatible (different type mapping, no ensure_ascii, different default hook protocol). The gopy port should implement json.dumps/json.loads as pure Go with a Python-compatible type dispatch: None->null, True/False->true/false, int->number, float->number, str->string, list->array, dict->object.