Conditional turn restriction support #3841

karenzshea · 2017-03-20T13:47:21Z

Issue

#3414

Initial pass at this will only support conditional turns not conditional speed limits.

So far, this branch contains updates for

conditional turn parsing by including an extra filter in osmium key parsing
writing found conditional turns to disk (and removing unconditional turn restrictions from being written to disk)
restriction file deserialization during update

Tasklist

Requirements / Relations

After this first pass is done, clean up tasks:

investigate how often applications of turn restrictions that expire immediately OR non-application of turn restrictions that will be shortly thereafter valid occur
normalize time range conditions to UTC during extraction

cc @TheMarex @oxidase

TheMarex · 2017-03-27T17:54:50Z

include/conditionals/extract-conditionals.hpp

+    ConditionalSpeedLimit speed_limit;
+};
+
+auto ParseGlobalArguments(int argc, char *argv[]);


These should probably not be exposed globally, because they are just there to parse CLI parameters. This is important for the conditionals tool, but not for anything just using the functions.

karenzshea · 2017-03-27T22:11:17Z

@oxidase @daniel-j-h If either of you have a time to look at what's going on with inclusion of Shapefile into cmake outside of just the BUILD_TOOLS building, I would be grateful! 🙏

In my last commit I made a flimsy pass at pulling out the find_package(Shapefile) results and exposing the paths to UPDATER_LIBRARIES and CONTRACTOR_LIBRARIES but I'm almost certainly missing something. Linking errors abound to all shapefile provided functions. Currently shapefil.h is only being included into the src/util/timezones.cpp file.

daniel-j-h · 2017-03-28T07:30:33Z

CMakeLists.txt

-  if(SHAPEFILE_FOUND)
-    add_dependency_includes(${LIBSHAPEFILE_LIBRARY})
+  if(NOT SHAPEFILE_FOUND)
+    message(FATAL_ERROR "Shapefile library (libshp-dev) was not found")


this is the same as find_package(P REQUIRED) (?)

yes, but i don't think it must be a requirement. better to have a warning and conditional compilation without time-based conditionals

daniel-j-h · 2017-03-28T07:31:01Z

CMakeLists.txt

@@ -682,7 +684,7 @@ target_link_libraries(osrm_store ${STORAGE_LIBRARIES})

 # BUILD_COMPONENTS
 add_executable(osrm-components src/tools/components.cpp $<TARGET_OBJECTS:UTIL>)
-target_link_libraries(osrm-components ${TBB_LIBRARIES} ${BOOST_BASE_LIBRARIES})
+target_link_libraries(osrm-components ${TBB_LIBRARIES} ${BOOST_BASE_LIBRARIES} ${LIBSHAPEFILE_LIBRARY})


why does components need the shapefile lib? transitive dep?

Yes, this should come from UTIL_LIBARIES eventually.

daniel-j-h · 2017-03-28T07:31:24Z

CMakeLists.txt

 target_link_libraries(osrm-extract osrm_extract ${Boost_PROGRAM_OPTIONS_LIBRARY})
 target_link_libraries(osrm-partition osrm_partition ${Boost_PROGRAM_OPTIONS_LIBRARY})
 target_link_libraries(osrm-customize osrm_customize ${Boost_PROGRAM_OPTIONS_LIBRARY})
 target_link_libraries(osrm-contract osrm_contract ${Boost_PROGRAM_OPTIONS_LIBRARY})
-target_link_libraries(osrm-routed osrm ${Boost_PROGRAM_OPTIONS_LIBRARY} ${OPTIONAL_SOCKET_LIBS} ${ZLIB_LIBRARY})
+target_link_libraries(osrm-routed osrm ${Boost_PROGRAM_OPTIONS_LIBRARY} ${LIBSHAPEFILE_LIBRARY} ${OPTIONAL_SOCKET_LIBS} ${ZLIB_LIBRARY})


same here, why does routed need shapefile lib? and datastore above?

util lib is not a library, so it is needed everywhere with $<TARGET_OBJECTS:UTIL>

it must be changed for sure

Same, this is a transitive dependency and should come from the UTIL_LIBRARIES variable.

oxidase · 2017-03-28T07:41:31Z

@karenzshea i have pushed changes to CMakeLists.txt, but actually i don't like to have Shapelib as a requirement. You can make optional by changing FATAL_ERROR to WARNING, define compilation flag -DSHAPELIB_FOUND and use it in the code later.

Also there is a problem with timezones.cpp in the "util lib" that is not really a library but a set of object files$<TARGET_OBJECTS:UTIL>. It requires to add everywhere ${LIBSHAPEFILE_LIBRARY}. I would suggest to move LoadLocalTimesRTree into a header file. It will make configuration with SHAPELIB_FOUND easier.

You can kick out my commit, it makes all build jobs red because libshp-dev is not installed. Only mason build should be green.

TheMarex · 2017-03-28T10:45:25Z

CMakeLists.txt

@@ -630,11 +635,15 @@ set(CUSTOMIZER_LIBRARIES
    ${MAYBE_RT_LIBRARY}
    ${MAYBE_COVERAGE_LIBRARIES})
 set(UPDATER_LIBRARIES
+    ${BZIP2_LIBRARIES}


BZIP2 and EXPAT should only be needed in EXTRACTOR after the header refactor.

Can you try to remove BZIP2 and EXPAND and see if it still compiles?

TheMarex · 2017-03-28T10:45:48Z

CMakeLists.txt

    ${TBB_LIBRARIES}
+    ${LIBSHAPEFILE_LIBRARY}


This should be moved to UTIL_LIBRARIES

TheMarex · 2017-03-28T10:47:05Z

CMakeLists.txt

@@ -682,7 +684,7 @@ target_link_libraries(osrm_store ${STORAGE_LIBRARIES})

 # BUILD_COMPONENTS
 add_executable(osrm-components src/tools/components.cpp $<TARGET_OBJECTS:UTIL>)
-target_link_libraries(osrm-components ${TBB_LIBRARIES} ${BOOST_BASE_LIBRARIES})
+target_link_libraries(osrm-components ${TBB_LIBRARIES} ${BOOST_BASE_LIBRARIES} ${LIBSHAPEFILE_LIBRARY})


Yes, this should come from UTIL_LIBARIES eventually.

TheMarex · 2017-03-28T10:47:44Z

CMakeLists.txt

 target_link_libraries(osrm-extract osrm_extract ${Boost_PROGRAM_OPTIONS_LIBRARY})
 target_link_libraries(osrm-partition osrm_partition ${Boost_PROGRAM_OPTIONS_LIBRARY})
 target_link_libraries(osrm-customize osrm_customize ${Boost_PROGRAM_OPTIONS_LIBRARY})
 target_link_libraries(osrm-contract osrm_contract ${Boost_PROGRAM_OPTIONS_LIBRARY})
-target_link_libraries(osrm-routed osrm ${Boost_PROGRAM_OPTIONS_LIBRARY} ${OPTIONAL_SOCKET_LIBS} ${ZLIB_LIBRARY})
+target_link_libraries(osrm-routed osrm ${Boost_PROGRAM_OPTIONS_LIBRARY} ${LIBSHAPEFILE_LIBRARY} ${OPTIONAL_SOCKET_LIBS} ${ZLIB_LIBRARY})


Same, this is a transitive dependency and should come from the UTIL_LIBRARIES variable.

TheMarex · 2017-03-28T11:00:33Z

CMakeLists.txt

@@ -521,6 +521,11 @@ else()
    ENDIF()
  ENDIF()

+  find_package(Shapefile) # package libshp-dev
+  if(NOT SHAPEFILE_FOUND)


We want to make the shapefile dependency optional. For this we need to ensure a few things:

We don't try to link against a missing library when it's not there

The code that uses shapefile is not compiled when the library is not found.

We get 1. by making the linker flags for shapefiles optional:

find_package(Shapefile) if (SHAPEFILE_FOUND) set(MAYBE_SHAPEFILE_LIBRARIES "${SHAPEFILE_LIBRARIES}") target_include_directories(UTIL PRIVATE ${LIBSHAPEFILE_INCLUDE_DIR}) else() set(MAYBE_SHAPEFILE_LIBRARIES "") endif()

This creates a variable that is just empty in case there is no shapefile library found, and only add the header include in case we find the library.

To get 2. we need to tell the C++ code whether it is allowed to use anything that references shapefiles. The typical technique to do that is use defines like #define FOO_BAR. Luckily we can pass these options straight to the compiler like:

g++ -D FOO_BAR test.cpp -o test

Conveniently cmake can generate those flags for us. Here is an example of how we do this to pass C++ the location of the test directory.

In our case that would be something like:

target_compile_definitions(UTIL PRIVATE COMPILE_DEFINITIONS ENABLE_SHAPEFILE)

daniel-j-h · 2017-04-04T09:27:32Z

include/conditionals/extract-conditionals.hpp

@@ -0,0 +1,384 @@
+#include "util/conditional_restrictions.hpp"


include guard missing

daniel-j-h · 2017-04-04T09:29:15Z

include/extractor/edge_based_graph_factory.hpp

 };
 #pragma pack(pop)
 static_assert(std::is_trivial<TurnIndexBlock>::value, "TurnIndexBlock is not trivial");
-static_assert(sizeof(TurnIndexBlock) == 24, "TurnIndexBlock is not packed correctly");
+static_assert(sizeof(TurnIndexBlock) < 24, "TurnIndexBlock is not packed correctly");


Why the change here? Byte size checks here should always check for equality.

daniel-j-h · 2017-04-04T09:31:01Z

include/extractor/extractor.hpp

@@ -56,14 +56,18 @@ class Extractor
  private:
    ExtractorConfig config;

+    std::vector<TurnRestriction> ParseOSMData(ScriptingEnvironment &scripting_environment,
+                                              const unsigned number_of_threads);


I think we can pull out the number of threads from the type's extractor config, no need to pass it in here

osrm-backend/include/extractor/extractor_config.hpp

Line 106 in e568b60

unsigned requested_num_threads;

Looks like a separate parameter is used here because at the call site extractor takes the min of a recommended_num_threads and the config provided num of threads.

Do we want to change this?

Ah okay then we're fine here!

daniel-j-h · 2017-04-04T09:32:07Z

include/extractor/io.hpp

+    const auto fingerprint = storage::io::FileReader::VerifyFingerprint;
+    storage::io::FileReader reader{path, fingerprint};
+
+    auto num_indices = reader.ReadElementCount32();


I would make all counts 64 bits in your new code already; see also #3848

daniel-j-h · 2017-04-04T09:36:25Z

include/extractor/io.hpp

+            reader.DeserializeVector(c.monthdays);
+        }
+        res.restriction.flags.is_only = is_only;
+        restrictions.push_back(res);


Above the code resizes the restrictions vector to size num_indices, but here you push_back into it.

I think what we want is either

To .reserve(num_indices) above (tells the vector we expect roughly num_incices elements) and then push_back here or

To .resize(num_indices) above and assign to restrictions[i] here

Think of reserve as a hint for vector to allocate memory more efficiently. In contrast, resize(n) adds n elements to the vector (default constructed).

daniel-j-h · 2017-04-04T09:37:57Z

include/extractor/io.hpp

+            reader.DeserializeVector(c.monthdays);
+        }
+        res.restriction.flags.is_only = is_only;
+        restrictions.push_back(res);


You can std::move(res) here to move the restriction container into the vector. At the moment we make a copy of res and then destroy res immediately after. Similar, you could put the InputRestrictionContainer above outside of the loop, otherwise we create an instance of it for every iteration.

daniel-j-h · 2017-04-04T09:38:50Z

include/extractor/io.hpp

+        reader.ReadInto(res.restriction.from);
+        reader.ReadInto(res.restriction.to);
+        reader.ReadInto(is_only);
+        for (auto &c : res.restriction.condition)


We never read into res.restriction.condition (?)

daniel-j-h · 2017-04-04T09:40:13Z

include/storage/io.hpp

@@ -104,6 +104,13 @@ class FileReader
        }
    }

+    void ReadInto(std::string &target)
+    {
+        auto len = ReadElementCount32();


Hm.. I would say 64 bit here for consistency. But this would also mean we require 8 bytes per string. These strings are osm strings, no? So it should be bound by 255 * 4 bytes (max. 4 bytes in utf-8 coded code point), and 16 bit should suffice?

I'm using this particular method for osm strings, but it's not unlikely that this method could be reused in the future for strings other than osm strings. Would a 16 bit size still work?

Depends how large the strings will be.

numeric_limits<uint8_t>::max(); // -> 255 numeric_limits<uint16_t>::max(); // -> 65535 numeric_limits<uint32_t>::max(); // -> 4294967295

OSM guarantees for a max of 256*4 bytes per string, so a uint16_t would be fine here. Other strings not coming from osm might need more.

daniel-j-h · 2017-04-04T09:43:38Z

include/util/timezones.hpp

+using local_time_t = std::pair<polygon_t, struct tm>;
+
+std::function<struct tm(const point_t &)> LoadLocalTimesRTree(const std::string &tz_shapes_filename,
+                                                              std::time_t utc_time);


dup. with timezoner class above?

daniel-j-h · 2017-04-18T13:58:38Z

CMakeLists.txt

@@ -524,6 +524,11 @@ else()
    ENDIF()
  ENDIF()

+  find_package(Shapefile) # package libshp-dev
+  if(NOT SHAPEFILE_FOUND)
+    message(FATAL_ERROR "Shapefile library (libshp-dev) was not found")


This is still on the todo list @karenzshea ?

daniel-j-h · 2017-04-18T14:27:33Z

include/extractor/serialization.hpp

+        }
+        restriction.flags.is_only = is_only;
+
+        restrictions.push_back(restriction);


If you take a look at the docs for push_back

http://en.cppreference.com/w/cpp/container/vector/push_back

you will see there are two overloads - one taking a const T& and the second taking a T&&. The first is what you call here: it will make a copy. But you know you don't need the restriction's contents anymore. In the next iteration loop it will be gone.

What you can do then, is to call std::move(restriction). The trick about move is, all it does is casting a T to a T&&. So what does it do? It allows the push_back(T&&) overload to be called instead, and this one is free to "steal" the restriction's content.

^ this is called move semantics; you can transfer "ownership" via std::move and avoid a copy this way.

(Note: a moved-from object is more or less empty, you must not access it again, as in: you can't e.g. move twice)

(Also please tell me if comments like this are helpful, or if I should just point out a move is missing here)

Arghhh sorry comments like these are very helpful 🙏 You already explained this to me once though, I was using an std::move before and then I must have rebased it away. Thanks for pointing out again!

daniel-j-h · 2017-04-18T14:28:29Z

include/extractor/serialization.hpp

+    writer.WriteOne(restriction.to);
+    writer.WriteOne(restriction.flags.is_only);
+    writer.WriteElementCount64(restriction.condition.size());
+    for (auto &c : restriction.condition)


Can c be a const&? We write from it to a file and the restriction object is already const.

daniel-j-h · 2017-04-18T14:31:29Z

include/util/timezones.hpp

@@ -0,0 +1,21 @@
+#ifndef TIMEZONES_HPP
+#define TIMEZONES_HPP


👍 for adding the include guard; for these generic'ish names maybe prefix with OSRM_ - the problem is, if we include a completely independent project and it also has a TIMEZONES_HPP guard, we will clash here.

daniel-j-h · 2017-04-18T14:32:56Z

src/extractor/extraction_containers.cpp

+            else
+            {
+                // save unconditional turn restriction to memory, for use in ebg later
+                unconditional_turn_restrictions.push_back(restriction_container.restriction);


Can we move here and above, too?

daniel-j-h · 2017-04-18T15:01:09Z

src/extractor/restriction_parser.cpp

+    }
+
+    // push back a copy of turn restriction
+    parsed_restrictions.push_back(boost::make_optional(restriction_container));


I was just about to say, can we make_optional(move(resr_cont)), and then checked optional's source:

http://www.boost.org/doc/libs/1_63_0/boost/optional/optional.hpp

Looks like the overload for make_optional(Args&& ...) is missing. It's in C++17's stdlib, though:

http://en.cppreference.com/w/cpp/utility/optional/make_optional

okay, bug, ticketed boostorg/optional 30

if you want to avoid a copy here you can do

optional<YourType> tmp{std::move(v)}; r.push_back(std::move(tmp));

I can do this, but requires removing TryParse's constness as a function

daniel-j-h · 2017-04-18T15:09:02Z

src/extractor/restriction_parser.cpp

+    // push back a copy of turn restriction
+    parsed_restrictions.push_back(boost::make_optional(restriction_container));
+
+    return std::move(parsed_restrictions);


The C++ guarantees for function-local variables to be moved out by default. If you add a move you penalize further optimization.

By which I mean

T fn() { T tmp; .. return tmp; // will be moved out if possible }

return move(tmp) would be bad for optimization here. Also there are situations in which you have to move:

struct S { T gn() { return tmp; // will _not_ move tmp out, as tmp is not a function-local variable } T tmp; };

(will crash if you cann gn() more than once, since a moved-from object is in a valid but unspecified state)

There are some articles out there, search for penalizing move, RVO, e.g.

https://www.ibm.com/developerworks/community/blogs/5894415f-be62-4bc0-81c5-3956e82276f3/entry/RVO_V_S_std_move?lang=en

daniel-j-h · 2017-04-18T15:10:24Z

src/extractor/scripting_environment_lua.cpp

+                    {
+                        for (boost::optional<InputRestrictionContainer> r : result_res)
+                        {
+                            resulting_restrictions.push_back(r);


for (T t : res)

will make a copy for every r in res; which is fine if you want to make a move(r) here

daniel-j-h · 2017-04-18T15:12:46Z

src/updater/updater.cpp

+    {
+        return hash_val(std::get<0>(t), std::get<1>(t));
+    }
+};


Where does hash_val come from? Also can you just use boost::hash they properly do the hash mixing part

http://www.boost.org/doc/libs/1_63_0/doc/html/hash.html

Comes from our own implementation of std_hash. Boost::hash implementation for tuples doesn't work across all the systems we test on #3962 (comment)

daniel-j-h · 2017-04-18T15:24:00Z

src/updater/updater.cpp

+    // TODO make this into a function
+    LookupTable<std::tuple<NodeID, NodeID>, NodeID> is_only_lookup;
+    std::unordered_set<std::tuple<NodeID, NodeID, NodeID>,
+                       std::hash<std::tuple<NodeID, NodeID, NodeID>>>


Here you'd need to change to boost::hash and it should magically just work

This doesn't work, boost::hash implementation for tuples doesn't work across all the systems we test on #3962 (comment)

oxidase · 2017-05-08T22:19:45Z

@karenzshea i have pushed a fix for msvs. please don't split tests into conditionals and non-conditionals: if there is no TZ information tm must fall back to the input UTC. This can be used to setup feature tests with conditionals without TZ shape files in both linux and windows builds.

EDIT: smth like

{
   std::vector<local_time_t> local_times;
   local_times.emplace_back(local_time_t{polygon_t{}, UTC time});
#ifdef ENABLE_SHAPEFILE
...
          return return local_times[0].second;
     };
#endif
  return [&local_times](const point_t &point) { return local_times[0].second; };


}

EDIT2: kicked out my commit: still Fatal Error C1060

oxidase · 2017-05-10T15:38:20Z

CMakeLists.txt

@@ -337,6 +337,7 @@ elseif(${CMAKE_CXX_COMPILER_ID} STREQUAL "MSVC")
  # using Visual Studio C++
  set(BOOST_COMPONENTS ${BOOST_COMPONENTS} zlib)
  set(CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS} /bigobj") # avoid compiler error C1128 from scripting_environment_lua.cpp
+  set(CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS} /Zm200")    # avoid compiler error C1060 due to Boost.Spirit on AppVeyor


O, this is leftover from msvc checks, please remove the line

TheMarex · 2017-05-10T18:32:47Z

@oxidase @karenzshea I really disagree about removing testing with a shapefile here. There should be no silent fallback to UTC, a shapefile should be a hard requirement for conditional turns.

If shipping the original shapefile is a problem we need to ship our own. Splitting the testing is ugly but can't be avoided with optional dependencies.

oxidase · 2017-05-10T19:47:24Z

@TheMarex i don't see any problem in including dummy TZ shapefile, but keeping 50Mb binary data is not good anyway because just shapefile loading in debug takes more than 5 seconds.
I added a dummy tz file with two "timzone" polygons generated from https://gist.github.com/oxidase/3122c705c9533f73fc63b01b22ad3ea8

mloskot · 2017-10-17T15:03:30Z

Does this introduce requirement of GeoJSON timezone file during contraction step of the processing flow?

Does this look like a valid way to disable the updater step during contraction?

osrm::ContractorConfig config;
config.updater_config.valid_now = 0; // disable updating of conditional turn restriction (requires timezone file)
...
osrm::contract(config);

Otherwise, it throws "Missing time zone geojson file" from the Timezoner ctor,

karenzshea · 2017-10-17T15:22:47Z

@mloskot This introduced an optional requirement of a GeoJSON timezone file.

Whether or not conditional turn restriction updating is applied during contraction/customization depends onthe presence of the .osrm.restrictions file, which contains the extracted conditional restrictions.

You can disable conditional turn restriction application by disabling the --parse-conditional-restrictions flag during osrm-extract, or removing the .osrm.restrictions file.

… parsing by default

karenzshea added the Work In Progress label Mar 20, 2017

karenzshea force-pushed the feature/condi-rice-turns branch 5 times, most recently from d1e39b1 to d0d296a Compare March 27, 2017 17:28

TheMarex reviewed Mar 27, 2017

View reviewed changes

karenzshea force-pushed the feature/condi-rice-turns branch 2 times, most recently from 535b4c3 to 488943f Compare March 27, 2017 22:11

daniel-j-h reviewed Mar 28, 2017

View reviewed changes

oxidase force-pushed the feature/condi-rice-turns branch 2 times, most recently from 4a5766d to 4b85dcf Compare March 28, 2017 08:32

TheMarex reviewed Mar 28, 2017

View reviewed changes

karenzshea force-pushed the feature/condi-rice-turns branch 2 times, most recently from 2c73307 to 2542b38 Compare April 3, 2017 15:08

daniel-j-h requested changes Apr 4, 2017

View reviewed changes

karenzshea force-pushed the feature/condi-rice-turns branch 2 times, most recently from 98cf68d to 9bdbdcd Compare April 13, 2017 18:34

karenzshea force-pushed the feature/condi-rice-turns branch from 9bdbdcd to 5771d99 Compare April 18, 2017 10:12

daniel-j-h requested changes Apr 18, 2017

View reviewed changes

karenzshea mentioned this pull request Apr 20, 2017

Removing writing of .timestamp file in extraction #3961

Closed

karenzshea force-pushed the feature/condi-rice-turns branch 7 times, most recently from 314da31 to b2bce1b Compare April 21, 2017 11:15

oxidase force-pushed the feature/condi-rice-turns branch from c7a84e6 to edca66e Compare May 9, 2017 06:44

oxidase added 2 commits May 9, 2017 14:09

Simplify Timezoner for windows build

701ac89

Split declaration and parsing parts for opening hours

cb7e7a4

oxidase force-pushed the feature/condi-rice-turns branch from edca66e to cb7e7a4 Compare May 9, 2017 12:09

karenzshea force-pushed the feature/condi-rice-turns branch from c83e9de to bbd2107 Compare May 10, 2017 10:57

adjust conditional tests to run without shapefiles

bbd2107

daniel-j-h approved these changes May 10, 2017

View reviewed changes

karenzshea force-pushed the feature/condi-rice-turns branch from fa219ce to cbcbaf2 Compare May 10, 2017 11:58

always include parse conditionals option

cbcbaf2

oxidase reviewed May 10, 2017

View reviewed changes

Adjust travis timeout

84b33ac

oxidase force-pushed the feature/condi-rice-turns branch from 4da0582 to a520cb0 Compare May 10, 2017 20:15

Added dummy TZ shapefile with test timezone polygons

a3e7d85

oxidase force-pushed the feature/condi-rice-turns branch from a520cb0 to a3e7d85 Compare May 10, 2017 20:16

TheMarex added Ready To Merge and removed Review - In feedback labels May 11, 2017

[skip ci] update changelog

c13e852

karenzshea merged commit 799a677 into master May 11, 2017

karenzshea deleted the feature/condi-rice-turns branch May 11, 2017 10:14

oxidase mentioned this pull request May 24, 2017

Support for conditional restriction and/or speed limits #3414

Closed

5 tasks

danpat mentioned this pull request Jul 3, 2017

Support for conditional access tags #4231

Open

mloskot mentioned this pull request Oct 17, 2017

osrm-extract --parse-conditional-restrictions=0 flag still generates restrictions #4620

Closed

mloskot referenced this pull request Oct 18, 2017

construct extract/contract configs to disable conditional restriction…

5dfb1cb

… parsing by default

HendrikLeuschner mentioned this pull request Dec 19, 2017

Route by supporting physical limitation of height and weight #4409

Merged

mjjbell mentioned this pull request Dec 8, 2021

Time dependent restrictions #6168

Closed

		@@ -0,0 +1,384 @@
		#include "util/conditional_restrictions.hpp"

Conditional turn restriction support #3841

Conditional turn restriction support #3841

Conversation

karenzshea commented Mar 20, 2017 • edited by TheMarex Loading

Issue

Tasklist

Requirements / Relations

Choose a reason for hiding this comment

karenzshea commented Mar 27, 2017

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

oxidase commented Mar 28, 2017

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

oxidase commented May 8, 2017 • edited Loading

Choose a reason for hiding this comment

TheMarex commented May 10, 2017

oxidase commented May 10, 2017 • edited Loading

mloskot commented Oct 17, 2017 • edited Loading

karenzshea commented Oct 17, 2017

karenzshea commented Mar 20, 2017 •

edited by TheMarex

Loading

oxidase commented May 8, 2017 •

edited

Loading

oxidase commented May 10, 2017 •

edited

Loading

mloskot commented Oct 17, 2017 •

edited

Loading