Probable restrictions #3361

gknisely · 2021-10-18T13:45:34Z

Issue

This PR consumes and utilizes probable route restrictions. The lowest probability allowed is set in the costing. Currently, only using probable restrictions that have a 100% probability.

Tasklist

Add tests
Update the changelog

Requirements / Relations

Link any requirements here. Other pull requests this PR is based on?

…probable. This is a hack for now and it should be restriction:probable

… 0 and ignore all probable restrictions.

…on exists

gknisely · 2021-10-18T13:50:04Z

lua/graph.lua

@@ -2091,8 +2091,14 @@ function rels_proc (kv, nokeys)
  end

  if (kv["type"] == "route" or kv["type"] == "restriction") then
+     if kv["restriction:probable"] then
+       if kv["restriction"] or kv["restriction:conditional"] then
+         kv["restriction:probable"] = nil


a user entered restriction always wins over a probable one

gknisely · 2021-10-18T13:50:53Z

src/sif/autocost.cc

+constexpr float kDefaultUseTolls = 0.5f;      // Default preference of using toll roads 0-1
+constexpr float kDefaultUseTracks = 0.f;      // Default preference of using tracks 0-1
+constexpr float kDefaultUseDistance = 0.f;    // Default preference of using distance vs time 0-1
+constexpr uint32_t kDefaultProbability = 100; // Default percentage of allowing probable restrictions


currently only care about restrictions with a probability of 100

gknisely · 2021-10-18T13:51:41Z

valhalla/sif/dynamiccost.h

@@ -410,6 +410,13 @@ class DynamicCost {
      // Iterate through the restrictions
      const EdgeLabel* first_pred = &pred;
      for (const auto& cr : restrictions) {
+        if (cr->type() == baldr::RestrictionType::kNoProbable ||
+            cr->type() == baldr::RestrictionType::kOnlyProbable) {
+          if (probability_ == 0 || probability_ > cr->probablity()) {


never use a probable restriction that is 0

Wouldn't that be the logical way to disable probable restrictions?

@purew yes that is exactly what this is doing. probability_ = 0 shuts off looking at any probable restrictions

Tiny nit, just saw the spelling issue with probablity -> probability

@purew 🦅 👁️

What if probabiity_==0 and cr->probability()==0? (This would be odd?)

@ktatterso probabiity_==0 would be true and we would just continue

@ktatterso I forgot to mention that we never have a complex restriction with a probability of 0. It means you have no faith in the probable restriction at all

@gknisely Thanks, I guess I meant the second part, what does cr->probability()==0 mean? And you just answered it. 🙂

kuanb · 2021-10-19T16:10:56Z

valhalla/sif/dynamiccost.h

@@ -907,6 +917,8 @@ class DynamicCost {
    destination_only_penalty_ = costing_options.destination_only_penalty();
    maneuver_penalty_ = costing_options.maneuver_penalty();

+    probability_ = costing_options.probability();


To re-confirm this is the logic that allows the probability threshold to be passed through in the payload under the costing_options key, right?

"costing_options": {"auto": {"probability": prob}}

Where prob is some values 0-100.

This means that, even though we default to 100, we can, when running Valhalla ourselves, pass in alternative thresholds when testing new PRRs/thresholds.

(cc @denmoroz)

@kuanb this is correct

purew · 2021-10-19T17:11:29Z

valhalla/baldr/complexrestriction.h

+   * Get the probability for the restriction.
+   * @return  Returns the probability for this restriction.
+   */
+  uint64_t probablity() const {


Is this a value between 0 and 100 that means percentage? If so could this be mentioned in doc-string and return type be set to uint8_t?

@purew all set.

purew · 2021-10-19T17:12:47Z

valhalla/mjolnir/complexrestrictionbuilder.h

+   * Set the probability for the restriction.
+   * @param  probability  probability for this restriction.
+   */
+  void set_probability(const uint64_t probability) {


Same comment as above, doc-string should mention what unit this is and input type should likely be shrunk to uint8_t.

@purew looks like a lot of the set and gets need to be fixed. Let me fix them.

@purew done

… uint16_t

purew

Left some more comments

purew · 2021-10-19T20:43:08Z

src/sif/autocost.cc

@@ -76,6 +78,7 @@ constexpr ranged_default_t<float> kUseTollsRange{0, kDefaultUseTolls, 1.0f};
 constexpr ranged_default_t<float> kUseDistanceRange{0, kDefaultUseDistance, 1.0f};
 constexpr ranged_default_t<float> kAutoHeightRange{0, kDefaultAutoHeight, 10.0f};
 constexpr ranged_default_t<float> kAutoWidthRange{0, kDefaultAutoWidth, 10.0f};
+constexpr ranged_default_t<float> kProbabilityRange{0, kDefaultProbability, 100};


The probability elsewhere is a uint8_t, should this kProbabilityRange be the same?

yes copy & paste fail

@purew I made this uint32_t to be consistent with the other options https://github.com/valhalla/valhalla/blob/master/proto/options.proto

purew · 2021-10-19T20:44:16Z

src/sif/autocost.cc

+    // probability
+    pbf_costing_options->set_probability(
+        kProbabilityRange(rapidjson::get_optional<uint32_t>(*json_costing_options, "/probability")
+                              .get_value_or(kDefaultProbability)));


Should we sanity check the option from the user, check that it is between 0 and 100 since it comes from an end-user?

@purew no need. this checks it for us constexpr ranged_default_t<uint8_t> kProbabilityRange{0, kDefaultProbability, 100};

purew · 2021-10-19T20:45:11Z

test/gurka/test_probable_restrictions.cc

+                                 {{"/date_time/type", std::to_string(GetParam())},
+                                  {"/date_time/value", "2021-04-02T19:20"}});
+  gurka::assert::raw::expect_path(result, {"AB", "BC", "CF"});
+}


Can you add a test where the probable restriction value is set as part of the request?

@purew sure. will do

@purew done

ktatterso · 2021-10-19T21:33:09Z

Seems like OSMRestriction should have a ctor that memset's itself to all 0's (like OSMWay). Otherwise the member data is unset.

…nt with proto.

dgearhart · 2021-10-20T03:42:00Z

proto/options.proto

@@ -132,6 +132,7 @@ message CostingOptions {
  optional bool include_hov2 = 74;
  optional bool include_hov3 = 75;
  optional bool exclude_cash_only_tolls = 76;
+  optional uint32 probability = 77;


name needs to be more descriptive - something like: restriction_probability

dgearhart · 2021-10-20T03:45:49Z

src/sif/autocost.cc

+constexpr float kDefaultUseTolls = 0.5f;      // Default preference of using toll roads 0-1
+constexpr float kDefaultUseTracks = 0.f;      // Default preference of using tracks 0-1
+constexpr float kDefaultUseDistance = 0.f;    // Default preference of using distance vs time 0-1
+constexpr uint32_t kDefaultProbability = 100; // Default percentage of allowing probable restrictions


naming again: kDefaultRestrictionProbability

dgearhart

made two naming change requests so it is more descriptive - other than that , it looks good to me

merkispavel

Check out my comment about default value/behaviour. I believe we can drop probability_ = 0 case from the code
#3361 (comment)

src/sif/autocost.cc

merkispavel

Looks good 👍

…striction_probability

merkispavel

🚢

gknisely · 2021-10-20T14:25:08Z

Going to RAD / QT test one more time due to all the changes

dgearhart

If RAD testing looks good then 🚢

gknisely · 2021-10-20T17:44:13Z

Seems like OSMRestriction should have a ctor that memset's itself to all 0's (like OSMWay). Otherwise the member data is unset.

@ktatterso going to have a small follow up pr for this as this has been like this for some time. Just don't want to mess up the functionality and this shouldn't, but you never know.

gknisely and others added 22 commits April 20, 2021 09:58

added support for restriction:motor_vehicle:probable and restriction:…

6776f23

…probable. This is a hack for now and it should be restriction:probable

added probability to the options

b7c51c0

added probability_ to equality operator.

91486a5

added logic to allow an inputted probability. We currently default to…

06c19b1

… 0 and ignore all probable restrictions.

added the accessor functions for the probability.

61bb7b1

added parsing logic for probable restrictions

e9623ed

added check for Probable restrictions.

a10db0c

added probability_ to complex restrictions.

04e1e25

added enums for probable restrictions.

cb8100f

added set_probability function.

ed52af1

updated sorting and added probability_

ac20b73

added logic for probable restrictions to restrict function.

36c2256

removed hack.

156ddc1

Update graph.lua

527af92

Merge remote-tracking branch 'origin/master' into probable_restrictions

ee38e3b

Merge remote-tracking branch 'origin/master' into probable_restrictions

09001f8

Merge remote-tracking branch 'origin' into probable_restrictions

ce7d00b

use only 100% Probability

17a7202

toss probable restriction if a conditional exists or normal restricti…

d4ea0a3

…on exists

lint

fa2372c

added tests.

31e0d32

not needed.

607a8ab

gknisely requested a review from merkispavel October 18, 2021 13:45

gknisely and others added 2 commits October 18, 2021 09:46

Merge branch 'master' into probable_restrictions

b767598

updated.

c418f13

gknisely commented Oct 18, 2021

View reviewed changes

kuanb reviewed Oct 19, 2021

View reviewed changes

purew reviewed Oct 19, 2021

View reviewed changes

gknisely and others added 4 commits October 19, 2021 14:18

cleanup. all should be uint8_t vs uint64_t except for modes. which is…

1465a6b

… uint16_t

Merge branch 'master' into probable_restrictions

f0957e5

clean up.

6e098e4

spelling typo

4f3db66

purew reviewed Oct 19, 2021

View reviewed changes

gknisely added 2 commits October 19, 2021 17:52

needs to be uint32_t and match proto. left as uint32_t to be consiste…

d4f5726

…nt with proto.

added more tests.

1c14e4b

dgearhart reviewed Oct 20, 2021

View reviewed changes

more descriptive name updates.

2631fdc

merkispavel requested changes Oct 20, 2021

View reviewed changes

src/sif/autocost.cc Outdated Show resolved Hide resolved

merkispavel previously approved these changes Oct 20, 2021

View reviewed changes

gknisely added 4 commits October 20, 2021 10:06

added test for no_left_turn @ probability=0 and probability is now re…

1d42dc0

…striction_probability

toss probable restriction if probability is 0

819af5a

refactored name. added more comments

9f8db7f

probability is now restriction_probability

e8e75f1

gknisely dismissed merkispavel’s stale review via e8e75f1 October 20, 2021 14:08

merkispavel approved these changes Oct 20, 2021

View reviewed changes

Merge branch 'master' into probable_restrictions

3d847e7

Merge branch 'master' into probable_restrictions

79dd3a7

dgearhart approved these changes Oct 20, 2021

View reviewed changes

gknisely merged commit feb722a into master Oct 21, 2021

gknisely added the hacktoberfest-accepted label Oct 21, 2021

gknisely deleted the probable_restrictions branch February 22, 2024 13:26

Probable restrictions #3361

Probable restrictions #3361

Conversation

gknisely commented Oct 18, 2021 • edited Loading

Issue

Tasklist

Requirements / Relations

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

purew Oct 19, 2021 • edited Loading

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

purew left a 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

ktatterso commented Oct 19, 2021

dgearhart Oct 20, 2021 • edited Loading

Choose a reason for hiding this comment

Choose a reason for hiding this comment

dgearhart left a comment

Choose a reason for hiding this comment

merkispavel left a comment • edited Loading

Choose a reason for hiding this comment

merkispavel left a comment

Choose a reason for hiding this comment

merkispavel left a comment

Choose a reason for hiding this comment

gknisely commented Oct 20, 2021

dgearhart left a comment

Choose a reason for hiding this comment

gknisely commented Oct 20, 2021

gknisely commented Oct 18, 2021 •

edited

Loading

purew Oct 19, 2021 •

edited

Loading

dgearhart Oct 20, 2021 •

edited

Loading

merkispavel left a comment •

edited

Loading