#4080)

* updated transifex

* Update CHANGELOG.md

* implement a fix for destonly edges around ferries, including the ability to not un-tag destonly edges further away from ferries, i.e. where there was a break to destonly continuity

* changelog

* new changelog block..

* revert most of the previous changes: we only don't penalize any dest_only while reclassifying if the connecting edge was dest_only AND the pred was destonly, but we still untag ANY destonly if it showed up in the final upclassified path; the main change is: also remove the connecting edge's destonly, which #1905 was missing; also fix a bug in forming the upclassified path where the initial label index could be wrong

0a4e828 removed the use values
`rail_connection` and `bus_connnection` and added the use values
`egress_connection` and `platform_connnection`.

The latter value contained a typo (three n) and the commit also didn't
update the documentation accordingly. This commit fixes these oversights.

* refactor: rename fo variable to file_op

* chore: facilitate spell-checking

* fix typo missed in previous PR

* ci: generalize changelog workflow to 'GitHub CI'

* ci: add spell-checking with typos

* Update CHANGELOG.md

---------

Co-authored-by: Kevin Kreiser <[email protected]>

* Remove unused variable

Apple clang generating this error:
 error: variable 'found' set but not used [-Werror,-Wunused-but-set-variable]

* Replace set with unordered_set

std::set is not defined in libc++ on MacOS XCode SDK 12 and is defined with a different type signature in XCode 13. There doesn't seem to be any need for the set to be ordered so this fixes the compilation errors

* Add maybe_unused attribute to found variable

This was providing a similar error as before:
 error: variable 'found' set but not used [-Werror,-Wunused-but-set-variable]
As the variable is only used when the logging level is set to trace.

* adding basic functionality of side cutoff

* added documentation

* added roundabout_exits to documentation

* applied clang-format

* added test

* updated changelog

* improved search filter order

* added comment in tests regarding driving side

* applied clang-tidy

* Updated description

Updated the description to be more understandable.

Co-authored-by: Kevin Kreiser <[email protected]>

* made street_side_cutoff optional in worker.cc

---------

Co-authored-by: = <[email protected]>
Co-authored-by: Kevin Kreiser <[email protected]>

* refactor valhalla_add_elevation a bit and move the mjolnir util to baldr

* add -j flag to all multithreaded executables

* changelog

* remove deprecated heder

* oops

* moved the argparse boilerplate code to a private header which all programs can share

* cxxopts entry not needed in cmake

* add missing midgard header

* some fixes

* huch, quite a logical mistake..

* few nits

* more fixes

* remove left-overs

* keep deduplication in the executable valhalla_add_elevation

* keep deduplication in the executable valhalla_add_elevation

* use __FILE__ to find the executable name; overhaul src/CMakeLists.txt, we need to put the ./src directory to target_include_directories()

* move get_tile_ids function into the executable

* add filesystem::stem() to return the file name without extension and adapt all executables

* remove just_gtfs official version

* use sorted feeds to speed up processing

* changelog

* use new no copy interfaces with different validity semantics

* lint changelog

* lint

* fix one of the test failures but the calendar part doesnt make sense..

* fix last bug, ready to ship

* fix yet another range bug. i see why c++20 made these first class citizens of stl

Co-authored-by: Kevin Kreiser <[email protected]>

* build database for parsed landmarks with tests

* remove unnecessary struct Landmark

* add bounding box query (TBC)

* fix small typo

* remove redundant code; code refactor and cleaning; add open db with different access

* add access mode in constructor, bounding box query TBC

* fix spatialite extension, finalize database builder

* formatting

* remove test func and fix loading spatilite by storing db_conn in the class

* update test with test fixtures

* move prepared statements to class member variables

* add Vacuum and Analyze; update insert func input param; code cleaning

* rename test to test_landmarks.cc

* remove unused imports

* code cleaning

* update interface

* tidy

* update constructor, remove unnecessary enum

* update connect_database func to use exceptions

* add release statements

* update Changelog

* simplify original class and prepare for pimpl

* finish pimpl

* formant

* changelog

* sudo install python dependencies on OSX

* well then.. install cython urgh

* huh, wrong python3 maybe, pre-commit depends on pyyaml 6.0 and installs without issues, 5.9 should have a wheel as well, but when downgrading pip decides to build from source instead.. urgh. last resort: pin pre-commit to a version which depends on pyyaml < 6.0, but eventually we'll need to do the conan switch to 2.0

* get rid of cython again..

… to 16 bytes) (#4214)

…es (#4204)

* coalesce the various gurka access tests into one

* add failing test

* add handling for :forward tags for all major modes

* lint

* update taginfo

* changelog

* lint

* we can delude ourselves that bikes are vehicles

…ags and LOG_LEVEL=WARN (#4197)

* implement callback and landmark parse function - first version

* implement landmark parser and test passed

* tidy

* add BuildLandmarkFromPBF function: parse landmarks from pbf and store them in database

* fix bug 'overwrite'

* tidy

* update Landmark type to enum of uint8_t with 15 popular amenity values, and update database part accordingly

* remove redundant func and test, code cleaning

* update LandmarkType to include only obvious landmarks for vehicle routing

* rename to landmark_builder

* move insertion of landmarks into node callback to have them directly inserted in to db after being parsed

* move landamark_callback to anonymous namespace

* add default values to LandmarkType

* update landmark_callback to parse landmark nodes only; update landmark test with border cases

* update test detail

* change default landmark name to empty string and remove defaults for lat and lng

* fix LandmarkDatabase constructoe, and fix rebase issues

* remove unnecessary tag 'landmark' in landmark_callback and update test

* update landmark_callback to not pass LandmarkDatabase as parameter in constructor

* code cleaning

* update BuildLandmarkFromPBF interface to take ptree for executable

* enable separate executable for landmrk builder

* enable valhalla_build_landmarks commandline

* fix typo

* remove comments

* remove struct Landmark and update interface and tests

* code cleaning in landmark tests

* change landmark database to only allow writes at creation; remove LandmarkType::null; code cleaning

* update CHANGELOG.md and description

* remove bus_station and biergarten in LandmarkType

* address code reviews

* add remarks for LandmarkType

* address code reviews

* update CHANGELOG.md

… key (#4224)

* [test/gurka/test_traffic.cc] sign-compare: int -> uint32_t

* [test/mapmatch.cc] sign-compare: size_t -> int

* [test/gurka/test_64bit_wayid.cc] sign-compare: int -> size_t

* [test/gurka/test_osrm_serializer.cc] sign-compare: int -> size_t

* [test/gurka/test_precision.cc] sign-compare: int -> size_t

* [test/gurka/test_time_dependent_tags.cc] sign-compare: int -> size_t

* [test/gurka/test_turn_lanes.cc] sign-compare: int -> size_t

* [test/http_tiles.cc] sign-compare: int -> size_t

* [test/shape_attributes.cc] sign-compare: int -> size_t

* [test/viterbi_search.cc] sign-compare: int -> size_t

* [test/gurka/test_traffic_smoothing.cc] sign-compare: int -> size_t

* [test/recover_shortcut.cc] sign-compare: int -> size_t

* [test/gurka/test_match.cc] sign-compare: int -> size_t & invalid/is_valid

* [test/gurka/test_route.cc] sign-compare: int -> size_t & invalid/is_valid

* [test/gurka/test_osrm_serializer.cc] Using rapidjson::SizeType

* [test/http_tiles.cc] Using std::filesystem

* [test/gurka/test_build_admin.cc] Removing unused variables

* [test/gurka/test_build_admin.cc] using std::filesystem

---------

Co-authored-by: Nils <[email protected]>

* fix crash in  timedistancebssmatrix.cc

* update changelog.md

* use protobuf's CMake config instead of CMake's own

* add todo

* ugly workaround

* manually find & link abseil to protobuf

* changelog

* comment out the dll line

* boil down to linking only the used abseil library

* fix: links on the optimized route API document

* docs: updated changelog

* skeleton for updates to the tile structure

* update AddLandmark

* fix bugs and code cleaning

* add string_to_landmark and vice versa in landmark header

* move stuff into baldr

* more structure to the test

* update Landmark back to struct

* update get_landmarks_by_bbox arguments order

* problem: sanity checks failed

* update Landmark::from_str to constructor

* fix encode and decode of lng and lat for landmark

* fix typo and argument names

* fix null byte bug by updating encoding of Landmark objects

* failing test

* some ramblings about how to get the data in ther

* back to fixed length encoding of Landmark lat and lng; update EdgeInfo::GetTaggedValues and EdgeInfo::GetTags()

* fix bug and allow more landmark tests

* tidy

* fix bug

* remove repeated imports

* update changelog

* use std::move and add to landmark test

* address code review

* complete test

* remove comments and optimize tests

* add a new line at the end of file

* remove fixed landmark in the test

---------

Co-authored-by: Kevin Kreiser <[email protected]>
Co-authored-by: Nils <[email protected]>

* Tyr osrm_serializer: Create bannerInstructions in step

* Tyr osrm_serializer: Make sure maneuver_type and turn_modifier isn't called twice

* Adds API flag to configure usage of bannerInstructions in OSRM serializer, default is false

* Tyr: Adds 'secondary' bannerInstructions

* Tyr: route_serializer_osrm apply clang-format

* Apply clang-format to worker.cc

* Tyr OSRM serializer removes unnecessary 'std::move' calls

* Tyr OSRM serializer break out intersections lane_indications function

* WIP: Tyr route_serializer_osrm - adds 'sub' lane instructions to bannerInstructions

* Tyr: OSRM serializer #lane_indications function pass turn_lane as pointer

* Tyr OSRM serializer: Makes sure, there is always a bannerInstructions in the final step

* Tyr OSRM serializer: Removes arrive_banner_instructions function

* Tyr OSRM serializer: Adds initial version of gurka test

* Tyr OSRM serializer: Apply clang-format

* Tyr OSRM serializer expand gurka test

* Tyr OSRM serialize improve gurka test to cover ref tag

* Tyr OSRM serializer: Creates additional bannerInstruction with 'sub' banner if the lanes come late in the step

* Tyr OSRM serializer: Adds lane_indications test

* Adds documentation on how to use the OSRM format with the optional banner_instructions

* Adds bannerInstructions enhancement to CHANGELOG

* Update proto/options.proto

Co-authored-by: Kevin Kreiser <[email protected]>

* Update docs/docs/api/turn-by-turn/api-reference.md

Co-authored-by: Kevin Kreiser <[email protected]>

* Tyr OSRM serializer: simplify lane_indications function signature

* Tyr OSRM serializer: Improve bannerInstructions - add exits, use only last lane information and pick better primary / secondary banners

* Tyr OSRM serializer: Add tests for motorway exit in bannerInstructions

* Tyr OSRM serializer: Adds degrees and driving_side to primary banner instruction for roundabouts and their exit steps

* Refactor test_osrm_serializer to use a test fixture class

* Small changes after code review

---------

Co-authored-by: Kevin Kreiser <[email protected]>

* update submodules which had new releases (unless it was a major version change)

* revert that patch

---------

Co-authored-by: Kevin Kreiser <[email protected]>

…kup/copy (#4255)

* add option to valhalla_build_extract to make an extract tar from an existing extract tar

* format and some cleanup

* changelog

* black

* amend tests

* deduplicate before writing to tar.. geojsons might catch many duplicates

* make the parent dir of the output tar file

* oops

* more oops

* one more ooops

* changelog

* skeleton for updates to the tile structure

* update AddLandmark

* fix bugs and code cleaning

* add string_to_landmark and vice versa in landmark header

* move stuff into baldr

* more structure to the test

* update Landmark back to struct

* update get_landmarks_by_bbox arguments order

* problem: sanity checks failed

* update Landmark::from_str to constructor

* fix encode and decode of lng and lat for landmark

* fix typo and argument names

* fix null byte bug by updating encoding of Landmark objects

* failing test

* some ramblings about how to get the data in ther

* back to fixed length encoding of Landmark lat and lng; update EdgeInfo::GetTaggedValues and EdgeInfo::GetTags()

* fix bug and allow more landmark tests

* tidy

* fix bug

* remove repeated imports

* update changelog

* use std::move and add to landmark test

* fix crash in  timedistancebssmatrix.cc (#4244)

* fix crash in  timedistancebssmatrix.cc

* update changelog.md

* fix up Cmake/protobuf disaster (#4207)

* use protobuf's CMake config instead of CMake's own

* add todo

* ugly workaround

* manually find & link abseil to protobuf

* changelog

* comment out the dll line

* boil down to linking only the used abseil library

* address code review

* complete test

* fix: links on the optimized route API document (#4260)

* fix: links on the optimized route API document

* docs: updated changelog

* skeleton for adding landmarks to the tiles

* more skeleton comments

* Update landmarks.cc

* more draft

* finish landmarks.cc and resolve compilation issues

* issue to be fixed

* bug: landmark added twice. the repeated landmark is always the last one to be added. NOTE: GraphTileBuilder::AddName cannot recognize the 'same' string. This bug is also correlated with landmark name size.

* fix bug and more tests

* fix bug of repeats and clean up code

* fix typo

* add a search cutoff for landmarks, and update tests

* add error test and use std::move in edgeinfo_offset_map_ update

* address code reviews and add comments

* update Changelog

* small update on const variable and tests

* add funciton description

* quick update

---------

Co-authored-by: Kevin Kreiser <[email protected]>
Co-authored-by: Patrick Qian <[email protected]>
Co-authored-by: Nils <[email protected]>
Co-authored-by: Pranav Chatly <[email protected]>

* removed clear_heading while sanitizing in loki_worker_t

#1802

* updated CHANGELOG.MD

#1802

---------

Co-authored-by: jehamannco <[email protected]>
Co-authored-by: Nils <[email protected]>

Co-authored-by: nilsnolde <[email protected]>

circleci changed how they do mac builds to support M1. now they only have one resource class (smaller than what we were using) for intel based builds so for the time being we have to use that: https://circleci.com/docs/using-macos/#supported-xcode-versions

* WIP: Add Singleton

* Change config singleton and refactor odin worker

* changed config singleton interface

* revert changes

* Remove toplevel references to config

* Clang tidy and format

* Add test

* Simplify test

* Remove additional config use

* Rename, simplify singleton and use other namespace

* clang-format

* include sorting

* test fix

* Update CHANGELOG.md

* Move config singleton/Remove unnessecary function

* Resolve naming collision

* format

* fix OneInstanceExisting test

* Fixed slipped file

* Add error handling and new exception to intercept

* Update CHANGELOG.md

---------

Co-authored-by: Nils <[email protected]>
Co-authored-by: johannes-no <[email protected]>
Co-authored-by: Kevin Kreiser <[email protected]>

…pronunciations (#4268)

* process default_language

* added LanguageTag

* added LanguageTag

* added logic to process default langs

* added default_languages

* added new GetTagTokens

* added lang lookups

* added language

* ported language logic

* more language updates

* not needed.

* rm jct

* format

* added lang check

* refactored to just have 1 linguistic record.

* added more entries

* kLanguage removed.  moved kNone to 5 to avoid versioning of the tiles

* refactored

* moved pronunciations and langs to maps.

* cleanup and added lang and pronunciation tags.

* Wales fix

* updated

* clang-tidy

* clean up

* updated.

* missed 2 changes.

* clean up

* more clean up

* added bridge

* pr clean up

* updated for new spell check

* more spell check

* Update CHANGELOG.md

* switched from 3 spaces to 2

* refactored to test_phonemes_w_langs.cc.  Reduced code duplication

---------

Co-authored-by: Greg Knisely <[email protected]>

…ibutes JSON (#4279)

* Initial code for storing elevation in Valhalla tiles.

* Add elevation to nodes and edgeinfo

* Added mjolnir code to populate node and edge elevation. Added code to
Valhalla route serializer and trace serializer to output elevation
(JSON) if elevation_interval is specified.

* Fix formatting

* Process elevation_interval JSON input

* Update api-reference.md

* Update api-reference.md

* Update api-reference.md

* Update api-reference.md

* Add to change log

* Address some PR comments.

* Update doc for elevation along route.

* Fix loop using size_t (was incorrect when addressing PR comments).

* Add to function level comments

* Use std::move to copy the encoded elevation vector.

* started elevation test

* more updates

* lend a hand on elevation tests

* finished up

* close the file

---------

Co-authored-by: Greg Knisely <[email protected]>
Co-authored-by: Kevin Kreiser <[email protected]>

* more sustainable way to work with protobuf in cmake

* let's piggyback on CMAKE_FIND_PACKAGE_PREFER_CONFIG

* see if arm has a jammy image available

* mason wont have clangformat for arm probably

* forget format for real th is time

* use newer vanilla ubuntu for arm, older versions of luajit are broken

* circle syntax

* syntax again

* try to share more and always use venv

* not sure where to get pip thought it came with venv

* really every time?

* git too

* stuff

* no format for now

* need sleep

* have to disable warnings to get it to build on arm

* try with sudo

* kick

* no sudo on arm?

* circle syntax

* skip format check for c++ on arm

* move sudo into the right places in the linux install script

* activate first so python is in its right mind

* use venv

* try to unbreak mac

* try prime server in /tmp and disable errors from bench

* fix isochrones on arm platforms

* lint

* mac already has venv

* already there

* unbreak the amd linux installs

* disable sanitizers so we can get a faster build

* fix astar shape comparisons

* lint

* more shape checks for arm

* actually fix the isochrone test, its off by 23 meters

* fix printing

* fix up landmark test to be aware of opposing edges

* lint

* simplify encode and decode of landmarks into tag values

* fix landmark test with slight buffer on bbox query

* sweet jesus

* changelog

* spelling

* update docker build to support arm as well at github actions. update circle to use env vars to use python virtual env

* try to remove virtual env

* circle lint

* more ci lint

* maybe this will work

* check if we need to install stuff

* use virtual env if in one..

* no file command

* on the release build we have to disable benchmark werror because of the newer compiler

* compiler too new with bugs!

* typo

* trying out multiplatform builds for docker

* typo

* try with sudo

* typo

* tag not working?

* oops

* update automatic workflow for docker publishing

* remove conan from windows build instructions

* clone cmf

* win path

* remove the god awful warnings from stdout

* screw ninja..

* git clone can't checkout single commits, only branch/tag

* weird.. missed boost-heap but linux didn't care. also remove old vcpkg deps file

* right, msvc does things differently than ninja

* upgrade vcpkg to newest master and update docs

* fix DST sensitive transit test

* fix tidy

…air (#4361)

* When adding a node label, add the edge's wayindex. When recovering the
shortest path, make sure the wayindex of the edge connecting the node
with its predecessor node matches the wayindex. This ensures the correct
edge on the shortest path is marked for reclassificiation. Prior to
this, multiple edges could become reclassified when there is more than
1 edge between a node pair.

* Change log

* Fix use of way_index when recovering shortest path.

* Add a test to make sure only the shorter edge between a node pair is
reclassified. Tested this against master and it fails. Succeeds on
the branch.

* formatting fix

* I hope it's that simple..

* changelog

* add test

* format

* Tidy

* Add test that demonstrates the wrong path can be recovered from the
ShortestPath method in ferry_connections.cc

* Fix the last_label_idx - make sure it is only set when an edge with the
target classification (or better) is present at the node being expanded.
Identify cases where no reclassification occurs (e.g,. islands with no
edges that meet the target classification) and log the locations.

* Add to change log

* If no path is found, always add to set of ferry locations with no path.

* Add test to make sure no edges are reclassified if the target
classification is not found.

* Looks like ReclassifyFerryConnections automatically sets the first edge
connecting to the ferry to primary. I added a test to make
sure no edges are reclassified if no edges with the target
classification are encountered. However, the initial ferry connection
edge is set to primary even if no edges with the target classification
are encountered in ShortestPath.

* Formatting

* remove some method declarations - not needed since only used in
ferry_connections.cc.

* More properly identify cases where no paths to/from ferries to target
classification roadways exist.

* formatting

* don't need to include unordered_set anymore.

* Remove the check if node_labels.size == 0. This is redundant now that we
check for a valid last_label_idx being set.

* Reclassify the first edge IF a connection to higher class roads is
found.

* Modify the nothing reclassified test to remove the first edge connected
to the ferry. No longer upclassed if no path to higher class edges is
found.

* copy tile pointer rather than mutate to avoid confusion

* make it work like the rest of the functions and comment it

* add a test for any reader functions that use tile cache args

* api doc strings

Co-authored-by: Evgen Bodunov <[email protected]>

Co-authored-by: Nils <[email protected]>

Co-authored-by: nilsnolde <[email protected]>

Co-authored-by: Nils <[email protected]>

* &  for verbose matrix response was kinda broken due to #4335 above

* add date_time array to slim matrix

* fix tid

* Nn matrix serializer fixes (#4385)

* src/tyr/matrix_serializer.cc is warning free

* Testing with true value and fixing test

* revert including date_time in non-verbose matrix

---------

Co-authored-by: Vicky Vergara <[email protected]>

…4413)

* Fix bug in build_trace on wrong edge_index assign

* Update Changelog

* Updated edge_index calculation logic

* Rollback edge_index type

---------

Co-authored-by: Mahdi Hasnat Siyam <[email protected]>

* save my work on the failing laptop

* refactor successful, all tests pass and the deadend bug is fixed..

* put the expansion direction into the template arguments

* rename vars

* do the memory retaining stuff as good as possible

* clean up other algos

* fix #4360 and the matrix-off-by-1-second issue

* put the dependencies back in place in cmakelists

* changelog

* format

* matrix expansion test also needed an update; good: we expand less edges:)

* oops, edgecost & transition cost were the same for both trees

* small update

* ha, what a shitshow.. tiny mis-ordering of statements messed it all up..

* last remaining bugs fixed

* cleanup

* initialize interrupt function with nullptr so using costmatrix by itself doesn't segfault

* re-add the additional testing scripts

* format

* increase the maximum threshold by 3% to account for slightly more expansion with deadend logic; still could/should be based on max cost of (which?) adjacency set

* after fixing the bugs, the expansion test has no changes

* revert

* make the PR files warning-free (#4410)

* address dave's review: change some of vicky's stuff around; add a gurka test for costmatrix to show that deadends are working now

* address review

* fix some warnings that crept in

* format

---------

Co-authored-by: Vicky Vergara <[email protected]>

* save my work on the failing laptop

* refactor successful, all tests pass and the deadend bug is fixed..

* put the expansion direction into the template arguments

* rename vars

* do the memory retaining stuff as good as possible

* clean up other algos

* fix #4360 and the matrix-off-by-1-second issue

* put the dependencies back in place in cmakelists

* changelog

* format

* matrix expansion test also needed an update; good: we expand less edges:)

* oops, edgecost & transition cost were the same for both trees

* small update

* ha, what a shitshow.. tiny mis-ordering of statements messed it all up..

* last remaining bugs fixed

* cleanup

* initialize interrupt function with nullptr so using costmatrix by itself doesn't segfault

* re-add the additional testing scripts

* format

* increase the maximum threshold by 3% to account for slightly more expansion with deadend logic; still could/should be based on max cost of (which?) adjacency set

* after fixing the bugs, the expansion test has no changes

* revert

* make the PR files warning-free (#4410)

* make reserverd memory for locations in CostMatrix configurable

* reorder

* address dave's review: change some of vicky's stuff around; add a gurka test for costmatrix to show that deadends are working now

* address review

* fix some warnings that crept in

* format

* changelog

---------

Co-authored-by: Vicky Vergara <[email protected]>

* Refactor EdgeLabel and derived classes to lower memory use. Fixed an
overflow of one of the 64 bit data members (lowers memory use by 8
bytes). Moved transition_cost and distance out of base EdgeLabel class.
Added a new dervied class, PathEdgeLabel, for use in path recosting.
Removed some unused Update methods. The base EdgeLabel class was reduced
to 40 bytes (was 56 bytes).

* Use the new PathEdgeLabel (derived class of EdgeLabel) for
LabelCallbacks and path recosting.

* Update calls to EdgeLabel constructor and methods.

* Add unit tests to validate EdgeLabel size (and its derived classes).

* Update EdgeLabel use.

* Update astar_bss to use BDEdgeLabel (as do all other A* methods).

* formatting

* Move distance to end of BDEdgeLabel. This leaves 4 bytes spare before
the size would increase.

* update change log

* Derive MMEdgeLabel from EdgeLabel (rather than PathEdgeLabel).
Reorganize distance data member to lower the size by 8 bytes (now
matches prior size so no size increase for MMEdgeLabel anymore).

* Actually MMEdgeLabel size did decrease to 72 bytes (was 80)!

* formatting (again)

* Move EdgeLabel size tests to double_bucket_queue test so that a new test
program does not need to be generated.

* format fix

* save my work on the failing laptop

* refactor successful, all tests pass and the deadend bug is fixed..

* put the expansion direction into the template arguments

* rename vars

* do the memory retaining stuff as good as possible

* clean up other algos

* fix #4360 and the matrix-off-by-1-second issue

* put the dependencies back in place in cmakelists

* changelog

* format

* matrix expansion test also needed an update; good: we expand less edges:)

* oops, edgecost & transition cost were the same for both trees

* small update

* ha, what a shitshow.. tiny mis-ordering of statements messed it all up..

* last remaining bugs fixed

* cleanup

* initialize interrupt function with nullptr so using costmatrix by itself doesn't segfault

* re-add the additional testing scripts

* format

* increase the maximum threshold by 3% to account for slightly more expansion with deadend logic; still could/should be based on max cost of (which?) adjacency set

* after fixing the bugs, the expansion test has no changes

* revert

* make the PR files warning-free (#4410)

* address dave's review: change some of vicky's stuff around; add a gurka test for costmatrix to show that deadends are working now

* address review

* fix some warnings that crept in

* format

* pure refactor to make room for the geometries in costmatrix

* add no_shape option for shape_format

* all in there incl serialization; now only needs the logic to actually put together the shape

* geometry works now; spent wayyyy too much time figuring out why trivial routes won't work: costmatrix has a bug where it finds the wrong direction of the single edge

* format

* band-aid the trivial route bug in costmatrix

* add extensive testing

* merge master and changelog

* docs

* changelog

* address review

* Apply suggestions from code review

Co-authored-by: Kevin Kreiser <[email protected]>

* addressed review

* address bug for tdmatrix when shape_format -ne 'none'

* tidy

* rename 'none' to 'no_shape' in JSON API; fix changelog; merge master

* again tidy

* test broke after renaming the JSON API

* added warning when tdmatrix was used and shape_format was requested

---------

Co-authored-by: Vicky Vergara <[email protected]>
Co-authored-by: Kevin Kreiser <[email protected]>

restiction -> restriction

…AddEdgeInfo call (#4436)

* process default_language

* added LanguageTag

* added LanguageTag

* added logic to process default langs

* added default_languages

* added new GetTagTokens

* added lang lookups

* added language

* ported language logic

* more language updates

* not needed.

* rm jct

* format

* added lang check

* refactored to just have 1 linguistic record.

* added more entries

* kLanguage removed.  moved kNone to 5 to avoid versioning of the tiles

* refactored

* moved pronunciations and langs to maps.

* cleanup and added lang and pronunciation tags.

* Wales fix

* updated

* clang-tidy

* clean up

* updated.

* missed 2 changes.

* clean up

* more clean up

* added bridge

* pr clean up

* updated for new spell check

* more spell check

* Update CHANGELOG.md

* added test for graph filter

* added test for graph filter

* missing diff_names check.

* updated.

* Update src/mjolnir/graphfilter.cc

Co-authored-by: Kevin Kreiser <[email protected]>

* moved check to OpposingEdgeInfoMatches function

* format

* added comment and end at cleanup stage

* Update test_filter.cc

* more mainstream test setup (#4438)

* more mainstream test setup

* Update test_filter.cc

---------

Co-authored-by: Greg Knisely <[email protected]>

* created OpposingEdgeInfoDiffers since same logic is used in shortcutbuilder, graphfilter, and hierarchybuilder

* added/updated diff_names check

* Delete test/gurka/test_phonemes_ipa_w_langs.cc

* Delete test/gurka/test_phonemes_jeita_w_langs.cc

* Delete test/gurka/test_phonemes_katakana_w_langs.cc

* Delete test/gurka/test_phonemes_nt_sampa_w_langs.cc

* Update CHANGELOG.md

* moved OpposingEdgeInfoDiffers to graphtilebuilder

---------

Co-authored-by: Greg Knisely <[email protected]>
Co-authored-by: Kevin Kreiser <[email protected]>
Co-authored-by: Nils <[email protected]>

* updated country access overrides.

* Update CHANGELOG.md

* naive fix for greg's observation

* remove 0.5 again

* Update motorscootercost.cc

* Update CHANGELOG.md