util,assert: improve comparison performance #22258

BridgeAR · 2018-08-11T00:27:33Z

This improves the performance of util.isDeepStrictEqual and assert.deepStrictEqual. It is mainly just code split from #22197 to reduce the overall change. This commit can also be backported. It should have similar results on older V8 versions.

The complete key comparison got reworked and short cuts got in where possible.

I only benchmarked the strict mode as it should be the only important part. For the legacy mode the performance should stay similar to the one before (+ / - a few percent here and there). I used the benchmark from #22211.

Some of the functions used also got safer by instrumenting uncurryThis.

Benchmark

 assert/deepequal-buffer.js method='deepEqual' strict=1 len=100 n=20000                                                                  2.35 %       ±2.37%   ±3.16%   ±4.14%
 assert/deepequal-buffer.js method='deepEqual' strict=1 len=1000 n=20000                                                                -0.97 %       ±1.58%   ±2.11%   ±2.78%
 assert/deepequal-buffer.js method='notDeepEqual' strict=1 len=100 n=20000                                                      ***     13.62 %       ±4.99%   ±6.64%   ±8.65%
 assert/deepequal-buffer.js method='notDeepEqual' strict=1 len=1000 n=20000                                                       *      7.78 %       ±5.96%   ±7.94%  ±10.34%
 assert/deepequal-map.js method='deepEqual_mixed' strict=1 len=500 n=500                                                        ***     30.98 %       ±2.48%   ±3.32%   ±4.36%
 assert/deepequal-map.js method='deepEqual_objectOnly' strict=1 len=500 n=500                                                   ***     33.05 %       ±2.96%   ±3.96%   ±5.19%
 assert/deepequal-map.js method='deepEqual_primitiveOnly' strict=1 len=500 n=500                                                        -0.35 %       ±3.14%   ±4.18%   ±5.45%
 assert/deepequal-map.js method='notDeepEqual_mixed' strict=1 len=500 n=500                                                             -1.59 %       ±3.01%   ±4.00%   ±5.21%
 assert/deepequal-map.js method='notDeepEqual_objectOnly' strict=1 len=500 n=500                                                ***     30.12 %       ±2.38%   ±3.18%   ±4.16%
 assert/deepequal-map.js method='notDeepEqual_primitiveOnly' strict=1 len=500 n=500                                                     -0.64 %       ±3.29%   ±4.38%   ±5.70%
 assert/deepequal-object.js method='deepEqual' strict=1 size=100 n=5000                                                         ***     18.25 %       ±5.16%   ±6.90%   ±9.04%
 assert/deepequal-object.js method='deepEqual' strict=1 size=1000 n=5000                                                        ***     23.78 %       ±1.88%   ±2.50%   ±3.25%
 assert/deepequal-object.js method='deepEqual' strict=1 size=50000 n=5000                                                       ***     66.63 %       ±2.73%   ±3.64%   ±4.76%
 assert/deepequal-object.js method='notDeepEqual' strict=1 size=100 n=5000                                                        *      5.35 %       ±5.33%   ±7.09%   ±9.23%
 assert/deepequal-object.js method='notDeepEqual' strict=1 size=1000 n=5000                                                     ***     28.52 %       ±4.49%   ±5.98%   ±7.78%
 assert/deepequal-object.js method='notDeepEqual' strict=1 size=50000 n=5000                                                    ***     53.75 %       ±6.11%   ±8.23%  ±10.91%
 assert/deepequal-prims-and-objs-big-array-set.js method='deepEqual_Array' strict=1 len=20000 n=25 primitive='array'            ***     76.65 %       ±2.12%   ±2.83%   ±3.70%
 assert/deepequal-prims-and-objs-big-array-set.js method='deepEqual_Array' strict=1 len=20000 n=25 primitive='number'           ***     79.50 %       ±2.31%   ±3.09%   ±4.04%
 assert/deepequal-prims-and-objs-big-array-set.js method='deepEqual_Array' strict=1 len=20000 n=25 primitive='object'           ***     74.47 %       ±3.94%   ±5.29%   ±7.01%
 assert/deepequal-prims-and-objs-big-array-set.js method='deepEqual_Array' strict=1 len=20000 n=25 primitive='string'           ***     78.85 %       ±3.66%   ±4.87%   ±6.34%
 assert/deepequal-prims-and-objs-big-array-set.js method='deepEqual_Set' strict=1 len=20000 n=25 primitive='array'                      -5.18 %       ±8.30%  ±11.05%  ±14.39%
 assert/deepequal-prims-and-objs-big-array-set.js method='deepEqual_Set' strict=1 len=20000 n=25 primitive='number'              **    -11.85 %       ±7.77%  ±10.35%  ±13.47%
 assert/deepequal-prims-and-objs-big-array-set.js method='deepEqual_Set' strict=1 len=20000 n=25 primitive='object'                     -1.18 %       ±7.44%   ±9.90%  ±12.89%
 assert/deepequal-prims-and-objs-big-array-set.js method='deepEqual_Set' strict=1 len=20000 n=25 primitive='string'              **    -11.27 %       ±8.17%  ±10.87%  ±14.15%
 assert/deepequal-prims-and-objs-big-array-set.js method='notDeepEqual_Array' strict=1 len=20000 n=25 primitive='array'         ***     80.53 %       ±4.14%   ±5.56%   ±7.34%
 assert/deepequal-prims-and-objs-big-array-set.js method='notDeepEqual_Array' strict=1 len=20000 n=25 primitive='number'        ***     77.87 %       ±4.93%   ±6.59%   ±8.65%
 assert/deepequal-prims-and-objs-big-array-set.js method='notDeepEqual_Array' strict=1 len=20000 n=25 primitive='object'        ***     76.43 %       ±3.70%   ±4.97%   ±6.54%
 assert/deepequal-prims-and-objs-big-array-set.js method='notDeepEqual_Array' strict=1 len=20000 n=25 primitive='string'        ***     77.90 %       ±1.73%   ±2.30%   ±3.00%
 assert/deepequal-prims-and-objs-big-array-set.js method='notDeepEqual_Set' strict=1 len=20000 n=25 primitive='array'           ***     29.18 %      ±10.95%  ±14.58%  ±18.99%
 assert/deepequal-prims-and-objs-big-array-set.js method='notDeepEqual_Set' strict=1 len=20000 n=25 primitive='number'          ***     26.29 %       ±8.99%  ±11.96%  ±15.58%
 assert/deepequal-prims-and-objs-big-array-set.js method='notDeepEqual_Set' strict=1 len=20000 n=25 primitive='object'          ***     32.19 %       ±8.86%  ±11.79%  ±15.35%
 assert/deepequal-prims-and-objs-big-array-set.js method='notDeepEqual_Set' strict=1 len=20000 n=25 primitive='string'           **     17.01 %      ±10.21%  ±13.61%  ±17.76%
 assert/deepequal-prims-and-objs-big-loop.js method='deepEqual' strict=1 n=20000 primitive='array'                              ***     26.84 %       ±3.13%   ±4.18%   ±5.47%
 assert/deepequal-prims-and-objs-big-loop.js method='deepEqual' strict=1 n=20000 primitive='number'                             ***     26.15 %       ±4.25%   ±5.68%   ±7.45%
 assert/deepequal-prims-and-objs-big-loop.js method='deepEqual' strict=1 n=20000 primitive='object'                             ***     30.46 %       ±2.63%   ±3.50%   ±4.56%
 assert/deepequal-prims-and-objs-big-loop.js method='deepEqual' strict=1 n=20000 primitive='string'                             ***     26.75 %       ±3.93%   ±5.26%   ±6.91%
 assert/deepequal-prims-and-objs-big-loop.js method='notDeepEqual' strict=1 n=20000 primitive='array'                           ***     28.66 %       ±5.13%   ±6.87%   ±9.03%
 assert/deepequal-prims-and-objs-big-loop.js method='notDeepEqual' strict=1 n=20000 primitive='number'                          ***     30.91 %       ±3.72%   ±4.95%   ±6.46%
 assert/deepequal-prims-and-objs-big-loop.js method='notDeepEqual' strict=1 n=20000 primitive='object'                          ***     31.59 %       ±1.75%   ±2.33%   ±3.03%
 assert/deepequal-prims-and-objs-big-loop.js method='notDeepEqual' strict=1 n=20000 primitive='string'                          ***     27.50 %       ±5.25%   ±7.05%   ±9.30%
 assert/deepequal-set.js method='deepEqual_mixed' strict=1 len=500 n=500                                                        ***     36.67 %       ±2.79%   ±3.71%   ±4.83%
 assert/deepequal-set.js method='deepEqual_objectOnly' strict=1 len=500 n=500                                                   ***     38.16 %       ±2.41%   ±3.20%   ±4.17%
 assert/deepequal-set.js method='deepEqual_primitiveOnly' strict=1 len=500 n=500                                                        -2.61 %       ±2.99%   ±3.98%   ±5.18%
 assert/deepequal-set.js method='notDeepEqual_mixed' strict=1 len=500 n=500                                                             -3.18 %       ±4.08%   ±5.42%   ±7.06%
 assert/deepequal-set.js method='notDeepEqual_objectOnly' strict=1 len=500 n=500                                                ***     35.45 %       ±2.72%   ±3.64%   ±4.80%
 assert/deepequal-set.js method='notDeepEqual_primitiveOnly' strict=1 len=500 n=500                                                     -2.36 %       ±3.94%   ±5.25%   ±6.83%
 assert/deepequal-typedarrays.js len=100 method='deepEqual' strict=1 n=500 type='Float32Array'                 ***     19.90 %       ±4.56%  ±6.09%  ±7.98%
 assert/deepequal-typedarrays.js len=100 method='deepEqual' strict=1 n=500 type='Float64Array'                 ***     19.50 %       ±2.67%  ±3.56%  ±4.65%
 assert/deepequal-typedarrays.js len=100 method='deepEqual' strict=1 n=500 type='Int8Array'                    ***     60.17 %       ±5.53%  ±7.36%  ±9.60%
 assert/deepequal-typedarrays.js len=100 method='deepEqual' strict=1 n=500 type='Uint8Array'                   ***     62.95 %       ±5.42%  ±7.22%  ±9.40%
 assert/deepequal-typedarrays.js len=100 method='deepEqual' strict=1 n=500 type='Uint8ClampedArray'            ***     65.83 %       ±8.33% ±11.09% ±14.45%
 assert/deepequal-typedarrays.js len=100 method='notDeepEqual' strict=1 n=500 type='Float32Array'              ***     40.78 %       ±3.13%  ±4.20%  ±5.54%
 assert/deepequal-typedarrays.js len=100 method='notDeepEqual' strict=1 n=500 type='Float64Array'              ***     42.95 %       ±1.76%  ±2.34%  ±3.05%
 assert/deepequal-typedarrays.js len=100 method='notDeepEqual' strict=1 n=500 type='Int8Array'                 ***    164.55 %      ±11.26% ±15.02% ±19.61%
 assert/deepequal-typedarrays.js len=100 method='notDeepEqual' strict=1 n=500 type='Uint8Array'                ***    156.12 %       ±9.02% ±12.08% ±15.89%
 assert/deepequal-typedarrays.js len=100 method='notDeepEqual' strict=1 n=500 type='Uint8ClampedArray'         ***    157.07 %       ±9.25% ±12.37% ±16.22%
 assert/deepequal-typedarrays.js len=5000 method='deepEqual' strict=1 n=500 type='Float32Array'                         1.50 %       ±1.99%  ±2.64%  ±3.44%
 assert/deepequal-typedarrays.js len=5000 method='deepEqual' strict=1 n=500 type='Float64Array'                         1.57 %       ±1.85%  ±2.47%  ±3.22%
 assert/deepequal-typedarrays.js len=5000 method='deepEqual' strict=1 n=500 type='Int8Array'                           -0.58 %       ±1.76%  ±2.35%  ±3.07%
 assert/deepequal-typedarrays.js len=5000 method='deepEqual' strict=1 n=500 type='Uint8Array'                          -1.53 %       ±3.29%  ±4.41%  ±5.80%
 assert/deepequal-typedarrays.js len=5000 method='deepEqual' strict=1 n=500 type='Uint8ClampedArray'                    0.05 %       ±1.95%  ±2.60%  ±3.39%
 assert/deepequal-typedarrays.js len=5000 method='notDeepEqual' strict=1 n=500 type='Float32Array'             ***     37.63 %       ±4.42%  ±5.92%  ±7.78%
 assert/deepequal-typedarrays.js len=5000 method='notDeepEqual' strict=1 n=500 type='Float64Array'             ***     35.06 %       ±4.47%  ±5.99%  ±7.89%
 assert/deepequal-typedarrays.js len=5000 method='notDeepEqual' strict=1 n=500 type='Int8Array'                ***     41.41 %       ±2.58%  ±3.44%  ±4.52%
 assert/deepequal-typedarrays.js len=5000 method='notDeepEqual' strict=1 n=500 type='Uint8Array'               ***     39.14 %       ±4.10%  ±5.51%  ±7.28%
 assert/deepequal-typedarrays.js len=5000 method='notDeepEqual' strict=1 n=500 type='Uint8ClampedArray'        ***     41.83 %       ±2.61%  ±3.48%  ±4.53%

Checklist

make -j4 test (UNIX), or vcbuild test (Windows) passes
tests and/or benchmarks are included
documentation is changed or added
commit message follows commit guidelines

nodejs-github-bot · 2018-08-11T00:27:34Z

@BridgeAR build started: https://ci.nodejs.org/blue/organizations/jenkins/node-test-pull-request-lite-pipeline/detail/node-test-pull-request-lite-pipeline/514/pipeline

benjamingr · 2018-08-11T07:39:38Z

lib/internal/util/comparisons.js

@@ -5,30 +5,53 @@ const { isArrayBufferView } = require('internal/util/types');
 const { internalBinding } = require('internal/bootstrap/loaders');
 const { isDate, isMap, isRegExp, isSet } = internalBinding('types');

-function objectToString(o) {
-  return Object.prototype.toString.call(o);
+const ReflectApply = Reflect.apply;


FWIW as a reviewer I'd still expect the "tamper safe"ing to be in a different PR. That said, changing it now here would be very frustrating to do likely so I'm fine with it here - just future personal preference.

I'll see if I move those out due to this becoming semver-major.

I only changed the single part as it is indeed more work to pull these changes out.

benjamingr · 2018-08-11T07:42:49Z

lib/internal/util/comparisons.js

-// barrier including the Uint8Array operation takes the advantage of the faster
-// binary compare otherwise. The break even point was at about 300 characters.
-function areSimilarTypedArrays(a, b, max) {
+function areSimilarFloatArrays(a, b) {
  const len = a.byteLength;


Is the caching of a.byteLength still helpful/needed?

No, but it should not hurt either?

Then can you remove it :)?

benjamingr · 2018-08-11T07:45:28Z

lib/internal/util/comparisons.js

-    // Skip testing the part below and continue with the keyCheck.
-    return keyCheck(val1, val2, true, memos);
+    }
+    // Fast path for non sparse arrays (no key comparison for indices


How does this if check assert this is a fast path and the array is not sparse?

From what I can tell:

this just checks that all the keys are indexed properties and that the last key is the last index

it relies on the Object.keys key ordering in V8 - AFAIK there is no guarantee on keys but there is on the order of other stuff e.g. getOwnPropertyNames (I am fine with this and just tracking it in case it changes in V8).

It fails with setting negative indices, for example: var o = {}; o[-1] = 3; o[1] = 5; o.length = 2. We should add a test for this particular case. For what it's worth I'm fine with that case being slow since it's uncommon and sparse arrays are really confusing perf-wise anyway.

Alternatively, we can see if we can check it directly from V8 - something like UseSparseVariant - or use %HasComplexElements

The keys are ordered by spec as follows:

The keys that are integer indices, in ascending numeric order.

All string keys, in the order in which they were added to the object.

All symbol keys, in the order in which they were added to the object.

The order makes this possible: Object.keys(arr)[arr.length - 1] === '' + (arr.length - 1). By checking the key at arr.length - 1 it should be the same key as the key length. Otherwise the key would have a position lower than that. If Object.keys(arr).length === arr.length on top of that, we also know that the array has no extra keys besides potentially symbols.

@BridgeAR as I said above as far as I know OrdinaryOwnPropertyKeys is used in .getOwnPropertyNames but not in .keys - that said I don't particularly mind it since I don't see a future in which V8 changes the iteration order of objects anyway since it's so heavily relied on.

I think negative indices are not a problem since they're not technically indices from the spec PoV and would be enumerated elsewhere.

Upon further inspection - it looks like this works fine for all the cases - care to add a comment?

By spec Object.keys calls EnumerableOwnPropertyNames and that calls OwnPropertyKeys and that calls OdinaryOwnPropertyKeys :)

Neat, I wonder when it changed - I remember this discussion when we did the io changes for ES2015 and it was guaranteed for one and not the other. Good to know it's a guarantee now - thanks.

benjamingr · 2018-08-11T07:56:51Z

lib/internal/util/comparisons.js

  }
  if (isDate(val1)) {
-    if (val1.getTime() !== val2.getTime()) {
+    if (getTime(val1) !== getTime(val2)) {


If someone subclasses Date and they have an overridden getTime() would this work?

If not, I am still +1 for it, but it would require this to be a very explicit semver-major.

I'll change it to a separate commit.

benjamingr · 2018-08-11T08:01:06Z

lib/internal/util/comparisons.js

+    if (keys.length !== objectKeys(val2).length) {
+      return false;
+    }
+    if (keys.length === val1.length) {


Nice changes in isArrayBufferView!

benjamingr · 2018-08-11T08:03:22Z

lib/internal/util/comparisons.js

+      return areSimilarTypedArrays(val1, val2);
+    }
+    if (isDate(val1) && isDate(val2)) {
+      return getTime(val1) === getTime(val2);


Ditto about getTime with a subclass (again +1 on the change, just making sure it's semver-major)

benjamingr · 2018-08-11T08:05:44Z

lib/internal/util/comparisons.js

+  }
+
+  // Cheap key test
+  let i = 0;


Where is i used outside of the loop here? It confused me since I expected it to be used at some capacity elsewhere. I don't think making i local to the loop would hurt performance.

It is used in two loops. I could define a second variable but this seemed nicer.

I find it a bit more confusing than defining it in the loop for the reason above but don't feel strongly about it - so if you think it's nicer I concede.

benjamingr · 2018-08-11T08:07:35Z

lib/internal/util/comparisons.js

@@ -451,10 +526,7 @@ function mapHasEqualEntry(set, map, key1, item1, strict, memo) {
 }

 function mapEquiv(a, b, strict, memo) {
-  if (a.size !== b.size)


Why did you inline the size checks rather than have them in mapEquiv (and setEquiv)?

I moved them to the beginning of the checks. Now, sets and maps will fail early in case the size does not match. Earlier they had significantly more work to reach this check.

benjamingr · 2018-08-11T08:08:14Z

lib/internal/util/comparisons.js

@@ -489,35 +561,44 @@ function mapEquiv(a, b, strict, memo) {
  return true;
 }

-function objEquiv(a, b, strict, keys, memos) {
+function objEquiv(a, b, strict, keys, memos, iterator) {


Not sure about naming it iterator but very much in favour of this change.

I am open to suggestions.

I'm not sure I like iteratorKind or iterationType are much better to be honest.

BridgeAR · 2018-08-11T09:01:09Z

test/parallel/test-assert-deep.js

@@ -890,3 +890,10 @@ assert.deepStrictEqual(obj1, obj2);
  );
  util.inspect.defaultOptions = tmp;
 }
+
+// Basic array out of bounds check.


This is already covered.

benjamingr

Left a bunch of nits I'd love to see resolved before landing but changes look good enough to not generate friction.

benjamingr

Actually looking at my review again I want to make sure that sparse arrays with negative indices work. I just did a code search on GitHub and found some code using that pattern which made me nauseous but reconsider landing this without verification this case works.

benjamingr · 2018-08-11T08:12:57Z

The semver-major is also preemptive, I'm fine with landing as semver-patch or semver-minor with a citgm run and a TSC member sign off.

benjamingr

Ok, I checked it out and am convinced this works + @BridgeAR explained the part I wasn't certain about with negative array indices.

I'm fine with this landing as-is as semver-major or with the getDate changes reverted as semver-patch.

(In case I haven't said it before - very nice changes!)

BridgeAR · 2018-08-11T11:17:23Z

I just reverted the getTime change, so it should be a patch again.

BridgeAR · 2018-08-11T11:22:15Z

@benjamingr thanks a lot for the thorough review!

CI https://ci.nodejs.org/job/node-test-pull-request/16367/

This adds a smarter logic to compare object keys (including symbols) and it also skips the object key comparison for (typed) arrays, if possible. Besides that it adds a fast path for empty objects, arrays, sets and maps and fast paths for sets and maps with an unequal size. On top of that a few functions are now safer to call by using uncurryThis and by caching the actual function. Overall, this is a significant performance boost for comparisons.

BridgeAR · 2018-08-13T20:53:32Z

Rebased due to conflicts.

CI https://ci.nodejs.org/job/node-test-pull-request/16427/

@nodejs/util @nodejs/testing PTAL

Trott

LGTM if CI is green

benjamingr

Still LGTM as semver-patch with the getTime reverts.

BridgeAR · 2018-08-15T15:42:33Z

Build is green with https://ci.nodejs.org/job/node-test-commit/20531/ and linux one rerun https://ci.nodejs.org/job/node-test-commit-linuxone/3974/

This adds a smarter logic to compare object keys (including symbols) and it also skips the object key comparison for (typed) arrays, if possible. Besides that it adds a fast path for empty objects, arrays, sets and maps and fast paths for sets and maps with an unequal size. On top of that a few functions are now safer to call by using uncurryThis and by caching the actual function. Overall, this is a significant performance boost for comparisons. PR-URL: nodejs#22258 Reviewed-By: Benjamin Gruenbaum <[email protected]> Reviewed-By: Rich Trott <[email protected]>

BridgeAR · 2018-08-15T15:45:25Z

Landed in d164d9d 🎉

This adds a smarter logic to compare object keys (including symbols) and it also skips the object key comparison for (typed) arrays, if possible. Besides that it adds a fast path for empty objects, arrays, sets and maps and fast paths for sets and maps with an unequal size. On top of that a few functions are now safer to call by using uncurryThis and by caching the actual function. Overall, this is a significant performance boost for comparisons. PR-URL: #22258 Reviewed-By: Benjamin Gruenbaum <[email protected]> Reviewed-By: Rich Trott <[email protected]>

BridgeAR added the performance Issues and PRs related to the performance of Node.js. label Aug 11, 2018

BridgeAR requested review from mscdex and benjamingr August 11, 2018 00:27

nodejs-github-bot added the util Issues and PRs related to the built-in util module. label Aug 11, 2018

BridgeAR added the assert Issues and PRs related to the assert subsystem. label Aug 11, 2018

benjamingr reviewed Aug 11, 2018

View reviewed changes

benjamingr approved these changes Aug 11, 2018

View reviewed changes

benjamingr added the semver-major PRs that contain breaking changes and should be released in the next major version. label Aug 11, 2018

benjamingr requested changes Aug 11, 2018

View reviewed changes

benjamingr approved these changes Aug 11, 2018

View reviewed changes

BridgeAR removed the semver-major PRs that contain breaking changes and should be released in the next major version. label Aug 11, 2018

BridgeAR added the author ready PRs that have at least one approval, no pending requests for changes, and a CI started. label Aug 11, 2018

This comment has been minimized.

Sign in to view

BridgeAR force-pushed the improve-comparison-perf branch from c6f41bf to f86a91f Compare August 13, 2018 20:52

BridgeAR requested a review from mcollina August 13, 2018 20:54

BridgeAR requested review from jasnell and devsnek August 13, 2018 20:54

Trott approved these changes Aug 13, 2018

View reviewed changes

benjamingr approved these changes Aug 14, 2018

View reviewed changes

joyeecheung mentioned this pull request Aug 15, 2018

CI failures: 20180815 nodejs/reliability#13

Closed

BridgeAR closed this Aug 15, 2018

BridgeAR mentioned this pull request Aug 15, 2018

util,assert: improve comparison performance #22197

Closed

4 tasks

targos mentioned this pull request Sep 5, 2018

Release proposal: v10.10.0 #22716

Merged

This was referenced Sep 6, 2018

2018-09-06 Version 10.10.0 (Current) @targos beevelop/docker-android-nodejs#236

Closed

2018-09-06 Version 10.10.0 (Current) @targos beevelop/docker-nodejs#221

Closed

BridgeAR mentioned this pull request Dec 20, 2019

for ... of replacing for(;;) in library networking code? #31024

Closed

BridgeAR deleted the improve-comparison-perf branch January 20, 2020 11:36

util,assert: improve comparison performance #22258

util,assert: improve comparison performance #22258

Conversation

BridgeAR commented Aug 11, 2018 • edited Loading

Checklist

nodejs-github-bot commented Aug 11, 2018

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

benjamingr Aug 11, 2018 • edited Loading

Choose a reason for hiding this comment

BridgeAR Aug 11, 2018 • 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

benjamingr Aug 11, 2018 • 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

BridgeAR Aug 11, 2018 • 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

This comment was marked as resolved.

Choose a reason for hiding this comment

benjamingr left a comment

Choose a reason for hiding this comment

benjamingr left a comment

Choose a reason for hiding this comment

benjamingr commented Aug 11, 2018

benjamingr left a comment

Choose a reason for hiding this comment

BridgeAR commented Aug 11, 2018

BridgeAR commented Aug 11, 2018

This comment has been minimized.

BridgeAR commented Aug 13, 2018

Trott left a comment

Choose a reason for hiding this comment

benjamingr left a comment

Choose a reason for hiding this comment

BridgeAR commented Aug 15, 2018

BridgeAR commented Aug 15, 2018

BridgeAR commented Aug 11, 2018 •

edited

Loading

benjamingr Aug 11, 2018 •

edited

Loading

BridgeAR Aug 11, 2018 •

edited

Loading

benjamingr Aug 11, 2018 •

edited

Loading

BridgeAR Aug 11, 2018 •

edited

Loading