make hgv-only edges also driveable to include in the reclassification logic #3860
Conversation
|
Might very well be.. Like I said, didn't look into it that much. Wanted to do that tmrw with you guys. Just wanted to get a quick check if the tests at least pass. Not that it means too much, but at least that:) |
|
So, @kevinkreiser is right, this is too naive. We'll have to do the second pass he mentions, when we find that the first pass found a path that's excluding certain modes. So effectively we'll find a second path to up-class (or even more depending on how many modes are missing). |
… for checking the right access direction
nilsnolde
force-pushed
the
nn-hgv-only-reclassification
branch
from
f6acce3
to
f19f846
Compare
… for checking the right access direction
nilsnolde
force-pushed
the
nn-hgv-only-reclassification
branch
from
f19f846
to
9c2b8ca
Compare
…lhalla into nn-hgv-only-reclassification
…ber which modes were accessible in the resulting path(s)
|
This is finished. I'll comment more in the PR. |
| if (node == (dt == Options::arrive_by ? "targets" : "sources")) { | ||
| for (auto& loc : locations) { | ||
| loc.set_date_time(dt == Options::current ? "current" : options.date_time()); | ||
| } |
There was a problem hiding this comment.
just removing a few lines
| class FerryTest : public ::testing::TestWithParam<std::string> { | ||
| protected: |
There was a problem hiding this comment.
I made this a parameterized test class with all vehicular modes as parameters
|
I first committed the test with 79f6ea0, so it's easy to just run it with current master. |
|
Hm.. Wanted to run it on the test data of #3731 but on a 200 MB PBF took > 5 mins reclassifying ferry edges before I canceled it.. The whole planet takes that long with |
… attribute for mjolnir::Edge to remember access for _any_ vehicular mode
…hole (auto) access logic when constructing edges in mjolnir
| uint16_t fwd_access; | ||
| uint16_t rev_access; |
There was a problem hiding this comment.
I removed the driveableforward & driveablereverse booleans and replaced all logic depending on them with the access masks which are only set if there's no Use::EmergencyAccess and has access to a vehicular mode.
I guess these would fit still into the attributes bit field, now that there's 25 bits available, though then only 1 would be left (access has 12 bits in directed edges) and we'd need to re-write to this if we needed more attributes. So I'd leave it this way.
src/mjolnir/node_expander.cc
Outdated
| auto edge_itr = edges[node.start_of]; | ||
| auto edge = *edge_itr; | ||
| auto edge = *edges[node.start_of]; | ||
| uint16_t auto_fwd_access = edge.fwd_access & baldr::kAutoAccess; |
|
OSX failed on the |
fixes #3731
I'm not entirely sure if this is fine for all downstream logic.. it seems not too much depends on this directly or indirectly. If someone has a concern, please voice it.
Also, again, IMO including truck here is enough, no need to respect more modes like motorbike.