i might have gotten through drafting the part that seemed most complex to me,
unsure
region_mask = aligned_tails[idx:next_idx-1] <
aligned_offsets[idx+1:next_idx]
region_bounds = region_mask.nonzero()[:,0]
region_bounds += idx
next_op_ct = op_ct + region_bounds.shape[0] + 1
offset_length_tail_idx_ops[op_ct:next_op_ct-1,TAIL] =
aligned_tails[region_bounds]
offset_length_tail_idx_ops[next_op_ct-1,TAIL] =
aligned_end
region_bounds += 1
offset_length_tail_idx_ops[op_ct+1:next_op_ct,OFFSET] =
aligned_offsets[region_bounds]
offset_length_tail_idx_ops[op_ct,OFFSET] = aligned_start
offset_length_tail_idx_ops[op_ct:next_op_ct,OP] =
OP_FETCH | OP_PLACE
above draft is intended to produce a list of pages that do that consolidation i
was talking about, where fetches are merged into surrounding pages (by using
aligned bounds) and adjacent pages are merged into a single fetch (by comparing
non-overlapping bounds for difference)
[.. function not drafted fully yet, maybe 1/3rd done. next is listing all the
ranges to output to the user and updating the loop variables over unfetched
regions kinda. there's still a loop but the intent is to only loop over sparse
and cached file regions (to cache accessed sparse regions)
