smash case

[yakra]

• For every AltLabel in every point not matching a primary waypoint label (most of the time) in every valid region/highway combo in every .list, we smash case 3 times:

  1. .lower() to compare to list labels
  2. .upper() to add to labels_in_use
  3. .upper() to check membership in unused_alt_labels & remove/erase

  Removing this redundancy can gain us some efficiency and clean up the code pending hash tables for waypoint labels? #278.

• Similarly, for every valid region/highway combo, we smash case on all alt_route_names until we match route_entry.

  Removing this redundancy can gain us some efficiency and clean up the code pending Output log for unused altRouteNames #230.

To be addressed in tandem with #316.

• Handle that first and squash commits.

ToDo:

• .upper() all alt_labels at once during Waypoint construction
• .strip('+') here too
• populate unused alt labels: don't smash case or .strip('+') anymore
• clean up alt_labels handling in TravelerList ctor
• .upper() all alt_route_names at once during Route construction
• don't smash case in TravelerList ctor (one line above Note: deprecated route name)
• clean up DUPLICATE_LABEL datacheck
• similarly .lower() root during Route construction
• make sure not lowered again during root_hash insertion
• .lower() all roots at once during ConnectedRoute construction
• change .lower() to .upper() as appropriate (not roots) globally

---

Similarly:
Another idea that may help speed up .list processing:

• Store an upper_label pre-stripped of leading + or * characters for each Waypoint upon construction, to avoid having to repeat this process every time we iterate through a route for a valid region/highway combo.
  Maybe this is a bit rich for my blood though, if the eventual idea is to implement hash tables for waypoint labels? #278.

[yakra]

Theoretical C++ speedup Reading waypoints for all routes.

• The refactored strip, upper and unused_alt_labels insertion takes 0.1s single-threaded on BiggaTomato, but we did also get rid of some extraneous string copying.
• Same deal for DUPLICATE_LABEL datacheck and its all_route_labels set, with a bit more copying thrown in due to using the std::string-based upper.
• These memory copies probably won't amount to much, but I already have the data here so why not look at it.
• As expected, no noticeable difference.

[yakra]

Oh crap. Just realized this will break nmpfps.log.

• Maybe the thing to do is read in wpts normally & handle near-miss points as we do now, then do another pass thru waypoints to smash case.
• populate unused alt labels here too.

[jteresco]

If I'm understanding the issue, could we just upcase or downcase the existing nmpfps.log when such a change would go live?

[yakra]

I suppose we could, but I've already worked around that.
It was just something I hadn't planned for, but thankfully the fix was very simple.

[yakra]

No diffs Python <-> C++.
Only old <-> new diff is DUPLICATE_LABEL entries in datacheck.log are caps now, rather than lower case.
siteupdate.py just ran to completion on noreaster.
Python speed comparison is underway on lab1, just > 20% done.
Time to commit, squash, and start the C++ speed tests!

[yakra]

Python: Time spent Processing traveler list files on lab1 is down by just over a third.