python/same-type-conversion
| Rule | same-type-conversion |
|---|---|
| Pack | Python |
| Severity | warning |
| Fix | unsafe, with xpd check --fix --unsafe-fixes |
| Library versions | every version |
float(), int(), str() or bool() around a value that already has that type does nothing. Drop the call.
def summary(prices: pd.Series, day: date) -> tuple[str, int, float, bool]: return ( str(day.isoformat()), int(prices.nunique()), float(prices.std()), bool(prices.empty), ) return day.isoformat(), prices.nunique(), prices.std(), prices.emptyWhat the rule reports
Section titled “What the rule reports”The rule reports float(), int(), str() and bool() around a value that has that type
already. The table lists what it reports and what it leaves alone.
| Call | Reported around | Left alone |
|---|---|---|
float() |
mean, median, std, var, sem, skew, kurt, corr and autocorr without an axis, in a file that imports numpy or pandas. np.sqrt, np.exp, np.log, np.hypot, np.mean, np.median, np.percentile, np.quantile, np.average, np.interp, np.trapezoid, np.linalg.norm and the nan forms without an axis. A division with / |
A call with axis=, and a file that imports torch, jax, tensorflow, xarray, cupy, dask, mlx, polars or sympy |
int() |
nunique(), .size, .ndim and .shape[0] |
sum, count, argmax and np.count_nonzero, which return np.int64. len(x) and round(x), which ruff's RUF046 reports |
str() |
An f-string, and isoformat, strftime, format, join, strip, lower, upper, title, removeprefix, removesuffix, read_text, decode, hexdigest and as_posix |
A plain string, which ruff's UP018 reports |
bool() |
not x, x in y, x is y, isinstance, any, all, startswith, endswith, exists, is_file, is_dir, .empty and .is_unique |
A comparison such as a < b, which returns np.bool on NumPy values |
How to fix it
Section titled “How to fix it”xpd check --fix --unsafe-fixes applies the fixes of this rule. Check the result, because an unsafe fix can change what the code does.
The safe fix writes the value without int(), str() or bool(). With the unsafe fix, the value without float(), because the mean of a float32 array is an np.float32, which is not a float. float() around a division has no fix.
xpd applies the replacement with --fix, and the one for float() with --fix --unsafe-fixes.
Each call converts a value to the type it has. The table shows what Python returned on NumPy
2.5.5 and pandas 3.1.1, and ty 0.0.105 accepted return x.mean() in a function that returns
float.
| Expression | Type | isinstance of the type in the name |
|---|---|---|
a.mean(), s.median(), s.quantile(0.5), s.corr(s) |
np.float64 |
float: True |
np.sqrt(a[0]), np.median(a), np.percentile(a, 50), np.interp(1.5, a, a), np.linalg.norm(a) |
np.float64 |
float: True |
a.astype("float32").mean() |
np.float32 |
float: False |
s.nunique(), s.size, a.ndim, a.shape[0] |
int |
int: True |
(s > 1).sum(), s.count(), a.argmax(), np.count_nonzero(a) |
np.int64 |
int: False |
s.empty, s.is_unique |
bool |
bool: True |
(s > 1).any(), a[0] < 1 |
np.bool |
bool: False |
int() around sum, count and argmax does change the type, and ty rejects
return x.argmax() in a function that returns int, so the rule leaves those calls alone. An
np.int64 works as an index and in arithmetic without a conversion. Convert it once, where the
value is written to JSON or returned as an int, and not at every step on the way.
Message
Section titled “Message”xpd prints this message with the issue. A name that starts with $ stands for code from your file.
$VALUE is a float already, because np.float64 is a subclass of float. Drop the float() around it.Browse the other rules of the Python pack.