Page 69 of 238

Re: street name not capitalized

Posted: Thu Feb 27, 2014 8:45 am
by dbraughlr
kentsmith9 wrote:Why are these two an error?
"S De la Cruz Blvd" → error.
"N Via de las Cruces Hwy" → error.
1. Failure to capitalize "La" of "La Cruz".
2. Failure to capitalize "Las" of "Las Cruces".

Re: road junctions and crossovers

Posted: Thu Feb 27, 2014 10:27 am
by dbraughlr
berestovskyy wrote: No doubt those buckets will reduce number of comparisons, but add complexity. And still there are will be thousands of comparisons and the check still will be very slow.
- a few thousand comparisons - not hundreds of thousands. But if it is too slow, make the squares smaller.
When the squares are under 25 m², the algorithm approaches Ο(n) time.
berestovskyy wrote: So we're looking at quite slow and complex check here. But what's is the "goal"? Does this non-ground elevation affects TTS or routing?
I would be looking for drivable segments at the same elevation that touch or cross without forming a junction. These might look okay, but they do not route correctly.

Re: street name not capitalized

Posted: Thu Feb 27, 2014 6:34 pm
by dbraughlr
kentsmith9 wrote: I saw a discrepancy between your two statements that I could not identify the correct solution.
Are you now satisfied that there is no discrepancy between my statements?
"La Cruz" gets capitalized for the same reason that "Las Cruces" is capitalized.
kentsmith9 wrote: I did not see the official list linked to what letters are supposed to be allowed in each country.
Current exceptions for the USA for first word of name:
  • to
kentsmith9 wrote: I now understand the second one if 'las' is not on the accepted list. I would then understand the first one if 'de' was not on the original displayed list.
The main point of my post was to demonstrate that allowing a word to be lowercase necessarily will fail to detect when that word should be capitalized and that the compass direction (e.g., "W de Vanter Ave") must be excluded when evaluating what is the first letter of a name.

Note to sketch: We have one (quirky) example so far in the entire USA. Programming Validator to allow some lowercase exceptions carries that trade-off that an incorrectly lowercase usage is accepted just so the rare exception won't be flagged.

Re: street name not capitalized

Posted: Thu Feb 27, 2014 7:09 pm
by dbraughlr
sketch wrote:A Validator check for first letter notwithstanding initial compass directions is fine ...
I am saying that this is necessary to make check #94 complete - and that checking every word against a list cannot substitute for checking the first actual word alone. So we need a list of words that are ignored at the beginning of a name ("E", "N", "S", "W", "to"). A note that false positives are theoretically possible is fine by me.

Re: street name not capitalized

Posted: Thu Feb 27, 2014 9:43 pm
by dbraughlr
kentsmith9 wrote: Thanks for the list. Is this list agreed by the whole USA community? I am not questioning your accuracy,
No. Go right ahead. I just made up the list myself as I was replying. It is not canon or accurate.
kentsmith9 wrote: Maybe the problem is berestovskyy's message was for all countries and yours was only for the USA?

But if that is the case, in California we have plenty of Spanish roads, so if "la" and "de" are allowed in other countries, why not allow it in the US?
I cannot comment much on other countries.

In California, we have many road names derived from Spanish, but "La" and "El" are always capitalized; "Del" is capitalized when it is the first real word of the name.

Rules for other countries do not necessary apply to the US.
They are capitalized in Mexico, so why shouldn't they be capitalized in the USA?

Re: same andpoints

Posted: Fri Feb 28, 2014 8:10 pm
by dbraughlr
Please move these posts to that thread (which was for the two-way ramp problem).

drivable segment has a very acute turn

Posted: Thu Mar 13, 2014 2:34 pm
by dbraughlr
A bowtie U-turn is not too sharp. These turns probably can be detected as a turn from a one-way street to a one-way street with the same or similar name.

Spreading the nodes can lead to an incorrect "stay right" instruction for traffic continuing straight through the intersection or missing road reports for drivers making the U-turn.

Re: drivable segment has a very acute turn

Posted: Thu Mar 13, 2014 8:44 pm
by dbraughlr
berestovskyy wrote:Validator does not report too-sharp bow-ties.
...
Sure, Validator can look for similar street names, but I guess it will eliminate any use of this check, because the check basically was designed to detect situations like this...
Yes, this is just a transition from divided to undivided. I thought that it was a simple case for demonstration. I didn't know that you had logic to detect an intersection.

A transition like this is valid and should be permitted.
What error would ever be missed by permitting this specific configuration?

Re: drivable segment has a very acute turn

Posted: Thu Mar 13, 2014 9:58 pm
by dbraughlr
1. The U-turn must be allowed.
2. Widening the U-turn could give rise to "Stay right" instructions for traffic going straight through.
3. Moving the split will result in Missing Road MPs when drivers make the legal U-turn.

Regardless, the example you gave doesn't match the transition scenario at all and would be detected because of either:
  1. The junction does not have a two-way street splitting into two one-way streets departing at less than 60 degrees from the two-way and within 90 degrees of each other. Or,
  2. The names of the one-way lanes do not begin with the name of the two-way street.
An at-grade connector is a one-way street that is usually unnamed. The duplicate right turn on the two-way portion should be disallowed.

Re: drivable segment has a very acute turn

Posted: Fri Mar 14, 2014 5:37 am
by dbraughlr
As long as we are quoting the wiki:
NOTE: Be aware of the geometry used in the bow tie. If the angles are too steep (45 degrees or greater) a driver traveling along one of the one-way segments may be told to "turn" or "stay" at the intersection instead of receiving no instruction.