read_excel: call skiprows callback just once per row.

For callable skiprows arguments, the callback was being called several
times in every row.

Now, it is called just once and the result cached.

Author: salva

pandas/io/excel/_base.py

@@ -817,8 +817,16 @@ def _parse_sheet(
         dtype_backend: DtypeBackend | lib.NoDefault = lib.no_default,
         **kwds,
     ):
-        try:
 
+        if callable(skiprows):
+            # In order to avoid calling skiprows multiple times on
+            # every row, we just do it here and keep the resulting
+            # list for passing it down to the parser.
+            skiprows = [ix for ix in range(len(data)) if skiprows(ix)]
+            if len(skiprows) == 0:
+                skiprows = None
+
+        try:
             # header indexes reference rows after removing skiprows, so we
             # create an index map from the without-skiprows to the
             # original indexes.
@@ -828,9 +836,7 @@ def _parse_sheet(
                 ixmap = list(range(skiprows, len(data)))
             elif is_list_like(skiprows):
                 skiprows_set = set(cast(Sequence[int], skiprows))
-                ixmap = [ix for ix, _ in enumerate(data) if ix not in skiprows_set]
-            elif callable(skiprows):
-                ixmap = [ix for ix, _ in enumerate(data) if not skiprows(ix)]
+                ixmap = [ix for ix in range(len(data)) if ix not in skiprows_set]
             else:
                 raise ValueError(
                     "skiprows must be an integer or a list of integers"