Skip to content

JSON Date Handling 1.0 Regressions #30977

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
merged 7 commits into from
Jan 18, 2020
Merged
Show file tree
Hide file tree
Changes from all commits
Commits
File filter

Filter by extension

Filter by extension

Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
122 changes: 76 additions & 46 deletions pandas/_libs/src/ujson/python/objToJSON.c
Original file line number Diff line number Diff line change
Expand Up @@ -456,8 +456,8 @@ static char *PyDateTimeToIso(PyDateTime_Date *obj, NPY_DATETIMEUNIT base,
static char *PyDateTimeToIsoCallback(JSOBJ obj, JSONTypeContext *tc,
size_t *len) {

if (!PyDateTime_Check(obj)) {
PyErr_SetString(PyExc_TypeError, "Expected datetime object");
if (!PyDate_Check(obj)) {
PyErr_SetString(PyExc_TypeError, "Expected date object");
return NULL;
}

Expand All @@ -469,7 +469,7 @@ static npy_datetime PyDateTimeToEpoch(PyObject *obj, NPY_DATETIMEUNIT base) {
npy_datetimestruct dts;
int ret;

if (!PyDateTime_Check(obj)) {
if (!PyDate_Check(obj)) {
// TODO: raise TypeError
}
PyDateTime_Date *dt = (PyDateTime_Date *)obj;
Expand Down Expand Up @@ -1504,6 +1504,7 @@ char **NpyArr_encodeLabels(PyArrayObject *labels, PyObjectEncoder *enc,
char **ret;
char *dataptr, *cLabel;
int type_num;
NPY_DATETIMEUNIT base = enc->datetimeUnit;
PRINTMARK();

if (!labels) {
Expand Down Expand Up @@ -1541,60 +1542,85 @@ char **NpyArr_encodeLabels(PyArrayObject *labels, PyObjectEncoder *enc,
break;
}

// TODO: vectorized timedelta solution
if (enc->datetimeIso &&
(type_num == NPY_TIMEDELTA || PyDelta_Check(item))) {
PyObject *td = PyObject_CallFunction(cls_timedelta, "(O)", item);
if (td == NULL) {
Py_DECREF(item);
NpyArr_freeLabels(ret, num);
ret = 0;
break;
}

PyObject *iso = PyObject_CallMethod(td, "isoformat", NULL);
Py_DECREF(td);
if (iso == NULL) {
Py_DECREF(item);
NpyArr_freeLabels(ret, num);
ret = 0;
break;
}

cLabel = (char *)PyUnicode_AsUTF8(iso);
Py_DECREF(iso);
len = strlen(cLabel);
} else if (PyTypeNum_ISDATETIME(type_num)) {
NPY_DATETIMEUNIT base = enc->datetimeUnit;
npy_int64 longVal;
int is_datetimelike = 0;
npy_int64 nanosecVal;
if (PyTypeNum_ISDATETIME(type_num)) {
is_datetimelike = 1;
PyArray_VectorUnaryFunc *castfunc =
PyArray_GetCastFunc(PyArray_DescrFromType(type_num), NPY_INT64);
if (!castfunc) {
PyErr_Format(PyExc_ValueError,
"Cannot cast numpy dtype %d to long",
enc->npyType);
}
castfunc(dataptr, &longVal, 1, NULL, NULL);
if (enc->datetimeIso) {
cLabel = int64ToIso(longVal, base, &len);
castfunc(dataptr, &nanosecVal, 1, NULL, NULL);
} else if (PyDate_Check(item) || PyDelta_Check(item)) {
is_datetimelike = 1;
if (PyObject_HasAttrString(item, "value")) {
nanosecVal = get_long_attr(item, "value");
} else {
if (!scaleNanosecToUnit(&longVal, base)) {
// TODO: This gets hit but somehow doesn't cause errors
// need to clean up (elsewhere in module as well)
if (PyDelta_Check(item)) {
nanosecVal = total_seconds(item) *
1000000000LL; // nanoseconds per second
} else {
// datetime.* objects don't follow above rules
nanosecVal = PyDateTimeToEpoch(item, NPY_FR_ns);
}
cLabel = PyObject_Malloc(21); // 21 chars for int64
sprintf(cLabel, "%" NPY_INT64_FMT, longVal);
len = strlen(cLabel);
}
} else if (PyDateTime_Check(item) || PyDate_Check(item)) {
NPY_DATETIMEUNIT base = enc->datetimeUnit;
if (enc->datetimeIso) {
cLabel = PyDateTimeToIso((PyDateTime_Date *)item, base, &len);
}

if (is_datetimelike) {
if (nanosecVal == get_nat()) {
len = 5; // TODO: shouldn't require extra space for terminator
cLabel = PyObject_Malloc(len);
strncpy(cLabel, "null", len);
} else {
cLabel = PyObject_Malloc(21); // 21 chars for int64
sprintf(cLabel, "%" NPY_DATETIME_FMT,
PyDateTimeToEpoch(item, base));
len = strlen(cLabel);
if (enc->datetimeIso) {
// TODO: Vectorized Timedelta function
if ((type_num == NPY_TIMEDELTA) || (PyDelta_Check(item))) {
Copy link
Member Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

This entire timedelta block was mostly copy / paste with a few changes at the end to fit the memory assumptions; this will be blown away by #30903

PyObject *td =
PyObject_CallFunction(cls_timedelta, "(O)", item);
if (td == NULL) {
Py_DECREF(item);
NpyArr_freeLabels(ret, num);
ret = 0;
break;
}

PyObject *iso =
PyObject_CallMethod(td, "isoformat", NULL);
Py_DECREF(td);
if (iso == NULL) {
Py_DECREF(item);
NpyArr_freeLabels(ret, num);
ret = 0;
break;
}

len = strlen(PyUnicode_AsUTF8(iso));
cLabel = PyObject_Malloc(len + 1);
memcpy(cLabel, PyUnicode_AsUTF8(iso), len + 1);
Py_DECREF(iso);
} else {
if (type_num == NPY_DATETIME) {
cLabel = int64ToIso(nanosecVal, base, &len);
} else {
cLabel = PyDateTimeToIso((PyDateTime_Date *)item,
base, &len);
}
}
if (cLabel == NULL) {
Py_DECREF(item);
NpyArr_freeLabels(ret, num);
ret = 0;
break;
}
} else {
cLabel = PyObject_Malloc(21); // 21 chars for int64
sprintf(cLabel, "%" NPY_DATETIME_FMT,
NpyDateTimeToEpoch(nanosecVal, base));
len = strlen(cLabel);
}
}
} else { // Fallback to string representation
PyObject *str = PyObject_Str(item);
Expand All @@ -1615,6 +1641,10 @@ char **NpyArr_encodeLabels(PyArrayObject *labels, PyObjectEncoder *enc,
ret[i] = PyObject_Malloc(len + 1);
memcpy(ret[i], cLabel, len + 1);

if (is_datetimelike) {
Copy link
Member Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

This was a memory leak before; new structure makes it easier to clean this up

PyObject_Free(cLabel);
}

if (PyErr_Occurred()) {
NpyArr_freeLabels(ret, num);
ret = 0;
Expand Down
26 changes: 26 additions & 0 deletions pandas/tests/io/json/test_pandas.py
Original file line number Diff line number Diff line change
@@ -1,4 +1,5 @@
from collections import OrderedDict
import datetime
from datetime import timedelta
from io import StringIO
import json
Expand Down Expand Up @@ -810,6 +811,31 @@ def test_convert_dates(self):
result = read_json(json, typ="series")
tm.assert_series_equal(result, ts)

@pytest.mark.parametrize("date_format", ["epoch", "iso"])
@pytest.mark.parametrize("as_object", [True, False])
@pytest.mark.parametrize(
"date_typ", [datetime.date, datetime.datetime, pd.Timestamp]
)
def test_date_index_and_values(self, date_format, as_object, date_typ):
data = [date_typ(year=2020, month=1, day=1), pd.NaT]
if as_object:
data.append("a")

ser = pd.Series(data, index=data)
result = ser.to_json(date_format=date_format)

if date_format == "epoch":
expected = '{"1577836800000":1577836800000,"null":null}'
else:
expected = (
'{"2020-01-01T00:00:00.000Z":"2020-01-01T00:00:00.000Z","null":null}'
)

if as_object:
expected = expected.replace("}", ',"a":"a"}')

assert result == expected

@pytest.mark.parametrize(
"infer_word",
[
Expand Down