Address review: drop unused batched reader, harden pick fast path

- Remove Raw._get_windows (unused in this changeset; will land with its
  first consumer).
- _picks_to_idx integer fast path now rejects duplicate picks explicitly
  via np.unique (suggested in review), keeping semantics identical to the
  slow path.
- Add changelog fragment.

Author: bruAristimunha

doc/changes/dev/14212.bugfix.rst

@@ -0,0 +1,4 @@
+Sped up raw data access for workloads making many small reads:
+``Raw.get_data`` no longer materializes the full time axis when ``tmin`` and
+``tmax`` are unset, resolves ``picks=None`` without channel-name machinery,
+and ``_mult_cal_one`` applies gather/type-cast/calibration in a single pass.

mne/_fiff/pick.py

@@ -1337,15 +1337,16 @@ def _picks_to_idx(
     # Fast path: an integer ndarray with all values already in range needs no
     # copy or further checks. This matters for callers resolving picks on
     # every access (e.g., Raw.get_data in deep-learning training loops).
-    if (
-        picks.dtype.kind == "i"
-        and picks.size
-        and picks.min() >= 0
-        and picks.max() < n_chan
-    ):
-        if return_kind:
-            return picks, picked_ch_type_or_generic
-        return picks
+    if picks.dtype.kind == "i" and len(picks):
+        sorted_picks = np.unique(picks)
+        if (
+            len(sorted_picks) == len(picks)
+            and sorted_picks[0] >= 0
+            and sorted_picks[-1] < n_chan
+        ):
+            if return_kind:
+                return picks, picked_ch_type_or_generic
+            return picks
     picks = picks.astype(int)
 
     #

mne/io/base.py

@@ -877,37 +877,6 @@ def _parse_get_set_params(self, item):
 
         return sel, start, stop
 
-    def _get_windows(self, starts, width, *, out=None, sel=None):
-        """Read many equal-width windows with setup shared across them."""
-        starts = np.atleast_1d(np.asarray(starts, dtype=np.int64)).ravel()
-        width = int(width)
-        if width <= 0:
-            raise ValueError(f"width must be positive, got {width}")
-        n_times = self.n_times
-        bad = (starts < 0) | (starts + width > n_times)
-        if bad.any():
-            raise ValueError(
-                f"window out of bounds at index {int(np.flatnonzero(bad)[0])}"
-            )
-        n_out = self.info["nchan"] if sel is None else len(sel)
-        if out is None:
-            out = np.empty((len(starts), n_out, width), dtype=self._dtype)
-        elif out.shape != (len(starts), n_out, width):
-            raise ValueError(
-                f"out has shape {out.shape}, need {(len(starts), n_out, width)}"
-            )
-        elif out.dtype not in (np.float64, np.float32):
-            raise ValueError(
-                f"out dtype must be float64 or float32, got {out.dtype}"
-            )
-        for j, s0 in enumerate(starts):
-            self._read_segment(
-                start=int(s0), stop=int(s0) + width,
-                sel=sel,
-                data_buffer=out[j],
-            )
-        return out
-
     def __getitem__(self, item):
         """Get raw data and times.