view contrib/fuzz/jsonescapeu8fast.cc @ 50400:95acba2c29f6

encoding: avoid quadratic time complexity when json-encoding non-UTF8 strings

Apparently the code uses "+=" with a bytes object, which is linear-time, so the
whole encoding is quadratic-time. This patch makes us use a bytearray object,
instead, which has a(n amortized-)constant-time append operation.

The encoding is still not particularly fast, but at least a 10MB file
takes tens of seconds, not many hours to encode.

#include <Python.h>
#include <assert.h>
#include <stdlib.h>
#include <unistd.h>

#include "pyutil.h"

#include <iostream>
#include <string>
#include "FuzzedDataProvider.h"

extern "C" {

static PYCODETYPE *code;

extern "C" int LLVMFuzzerInitialize(int *argc, char ***argv)
{
	contrib::initpy(*argv[0]);
	code = (PYCODETYPE *)Py_CompileString(R"py(
try:
    parsers.jsonescapeu8fast(data, paranoid)
except Exception as e:
    pass
    # uncomment this print if you're editing this Python code
    # to debug failures.
    # print(e)
)py",
	                                      "fuzzer", Py_file_input);
	if (!code) {
		std::cerr << "failed to compile Python code!" << std::endl;
	}
	return 0;
}

int LLVMFuzzerTestOneInput(const uint8_t *Data, size_t Size)
{
	FuzzedDataProvider provider(Data, Size);
	bool paranoid = provider.ConsumeBool();
	std::string remainder = provider.ConsumeRemainingBytesAsString();

	PyObject *mtext = PyBytes_FromStringAndSize(
	    (const char *)remainder.c_str(), remainder.size());
	PyObject *locals = PyDict_New();
	PyDict_SetItemString(locals, "data", mtext);
	PyDict_SetItemString(locals, "paranoid", paranoid ? Py_True : Py_False);
	PyObject *res = PyEval_EvalCode(code, contrib::pyglobals(), locals);
	if (!res) {
		PyErr_Print();
	}
	Py_XDECREF(res);
	Py_DECREF(locals);
	Py_DECREF(mtext);
	return 0; // Non-zero return values are reserved for future use.
}
}