view rust/hg-core/src/utils/hg_path.rs @ 45986:cc759d3db1e8

copies-rust: leverage the immutability for efficient update The immutable OrdMap has a diff function. That function yield items for each difference (missing in minor; missing in major; present in both, but different). We reorganise the manifest merging code to use this new function. We gather the data we would need to inject into the `minor` or `major` dict and eventually update one of them. The semantic of the merge itself is kept as is. Why are we doing this? Last year we contributed a patch making that diff function quickly detect some of the identical subsection when comparing two OrdMap cloned from a common ancestors (because they point to the very same node in memory). As a result: - That diff function is very fast in most of our cases, - It is important to maximize the "common" part by minimising the amount of unnecessary changes we do in theses Map. This is why we gather update for both and update the one with the smaller update. In practice, this yield a massive speed up on all our slow cases. Some examples below.: Repo Cases Source-Rev Dest-Rev Old-Time New-Time Difference Factor ------------------------------------------------------------------------------------------------------------------------------------ pypy x0000_revs_x_added_0_copies d1defd0dc478 c9cb1334cc78 : 33.527067 s, 2.780174 s, -30.746893 s, ? 0.0829 netbeans x0000_revs_xx000_added_x000_copies 588c2d1ced70 1aad62e59ddd : killed(>120), 1.231869 s mozilla-central x00000_revs_x0000_added_x0000_copies 6832ae71433c 4c222a1d9a00 : killed(>120), 3.478747 s mozilla-try x0000_revs_xx000_added_x000_copies 89294cd501d9 7ccb2fc7ccb5 : 83.508590 s, 9.843481 s, -73.665109 s, ? 0.1179 mozilla-try x0000_revs_x0000_added_x0000_copies e928c65095ed e951f4ad123a : 55.079813 s, 1.465128 s, -53.614685 s, ? 0.0266 The performance compared to the Python code are now comparable (worst case are a score percent slower)" You an check the full table below for details. The big new is that, with this change, we are now faster filelog in most case, . Below is an highlight of some pretty nice win: Repo Case Source-Rev Dest-Rev filelog sidedata Difference Factor --------------------------------------------------------------------------------------------------------------------------------------- pypy x0000_revs_xx000_added_x000_copies 08ea3258278e d9fa043f30c0 : 1.354211 s, 0.252125 s, -1.102086 s, ? 0.186179 netbeans x000_revs_x000_added_x000_copies ff453e9fee32 411350406ec2 : 0.939593 s, 0.112534 s, -0.827059 s, ? 0.119769 netbeans x0000_revs_xx000_added_x000_copies 588c2d1ced70 1aad62e59ddd : 3.824967 s, 1.231869 s, -2.593098 s, ? 0.322060 mozilla-central x000_revs_x000_added_x000_copies 7c97034feb78 4407bd0c6330 : 1.142414 s, 0.102184 s, -1.040230 s, ? 0.089446 mozilla-central x0000_revs_xx000_added_0_copies 9eec5917337d 67118cc6dcad : 6.667504 s, 0.046220 s, -6.621284 s, ? 0.006932 mozilla-central x0000_revs_xx000_added_x000_copies f78c615a656c 96a38b690156 : 3.267274 s, 0.315271 s, -2.952003 s, ? 0.096494 mozilla-central x00000_revs_x0000_added_x0000_copies 6832ae71433c 4c222a1d9a00 : 16.038104 s, 3.478747 s, -12.559357 s, ? 0.216905 mozilla-central x00000_revs_x00000_added_x000_copies 76caed42cf7c 1daa622bbe42 : 20.639928 s, 4.766435 s, -15.873493 s, ? 0.230933 They they are a handful of slower case where we perform less well. A good Share of them are really pathological "service" merge in mozilla-try, but not all of them. Repo Case Source-Rev Dest-Rev filelog sidedata Difference Factor --------------------------------------------------------------------------------------------------------------------------------------- pypy x0000_revs_x_added_0_copies d1defd0dc478 c9cb1334cc78 : 0.001055 s, 2.780174 s, +2.779119 s, ? 2635.236019 mozilla-try x0000_revs_x_added_0_copies 63519bfd42ee a36a2a865d92 : 0.088715 s, 0.435381 s, +0.346666 s, ? 4.907637 mozilla-try x0000_revs_x_added_x_copies 9fe69ff0762d bcabf2a78927 : 0.080765 s, 0.415461 s, +0.334696 s, ? 5.144072 mozilla-try x0000_revs_xx000_added_x000_copies 89294cd501d9 7ccb2fc7ccb5 : 7.598509 s, 9.843481 s, +2.244972 s, ? 1.295449 mozilla-try x00000_revs_x_added_0_copies 6a320851d377 1ebb79acd503 : 0.092232 s, killed mozilla-try x00000_revs_x_added_x_copies 5173c4b6f97c 95d83ee7242d : 0.093892 s, killed mozilla-try x00000_revs_x000_added_x_copies 9126823d0e9c ca82787bb23c : 0.227503 s, killed mozilla-try x00000_revs_x00000_added_x0000_copies 1b661134e2ca 1ae03d022d6d : 21.145391 s, killed mozilla-try x00000_revs_x00000_added_x000_copies 9b2a99adc05e 8e29777b48e6 : 25.304164 s, killed Below are two different tables for full performance comparison - this changeset against the previous one (spoiler: it is much better) - this changeset against the python code (spoiler: still slower, but it gets more comparable) - this changeset against the filelog code (spoiler: better in many case, but not all) Repo Cases Source-Rev Dest-Rev Old-Time New-Time Difference Factor ------------------------------------------------------------------------------------------------------------------------------------ mercurial x_revs_x_added_0_copies ad6b123de1c7 39cfcef4f463 : 0.000049 s, 0.000049 s, +0.000000 s, ? 1.0000 mercurial x_revs_x_added_x_copies 2b1c78674230 0c1d10351869 : 0.000179 s, 0.000182 s, +0.000003 s, ? 1.0168 mercurial x000_revs_x000_added_x_copies 81f8ff2a9bf2 dd3267698d84 : 0.006494 s, 0.005872 s, -0.000622 s, ? 0.9042 pypy x_revs_x_added_0_copies aed021ee8ae8 099ed31b181b : 0.000339 s, 0.000229 s, -0.000110 s, ? 0.6755 pypy x_revs_x000_added_0_copies 4aa4e1f8e19a 359343b9ac0e : 0.000057 s, 0.000058 s, +0.000001 s, ? 1.0175 pypy x_revs_x_added_x_copies ac52eb7bbbb0 72e022663155 : 0.000299 s, 0.000148 s, -0.000151 s, ? 0.4950 pypy x_revs_x00_added_x_copies c3b14617fbd7 ace7255d9a26 : 0.001200 s, 0.001205 s, +0.000005 s, ? 1.0042 pypy x_revs_x000_added_x000_copies df6f7a526b60 a83dc6a2d56f : 0.025120 s, 0.025662 s, +0.000542 s, ? 1.0216 pypy x000_revs_xx00_added_0_copies 89a76aede314 2f22446ff07e : 0.506921 s, 0.080113 s, -0.426808 s, ? 0.1580 pypy x000_revs_x000_added_x_copies 8a3b5bfd266e 2c68e87c3efe : 1.272060 s, 0.153030 s, -1.119030 s, ? 0.1203 pypy x000_revs_x000_added_x000_copies 89a76aede314 7b3dda341c84 : 0.690941 s, 0.098774 s, -0.592167 s, ? 0.1430 pypy x0000_revs_x_added_0_copies d1defd0dc478 c9cb1334cc78 : 33.527067 s, 2.780174 s, -30.746893 s, ? 0.0829 pypy x0000_revs_xx000_added_0_copies bf2c629d0071 4ffed77c095c : 0.021970 s, 0.022218 s, +0.000248 s, ? 1.0113 pypy x0000_revs_xx000_added_x000_copies 08ea3258278e d9fa043f30c0 : 1.772094 s, 0.252125 s, -1.519969 s, ? 0.1423 netbeans x_revs_x_added_0_copies fb0955ffcbcd a01e9239f9e7 : 0.000185 s, 0.000186 s, +0.000001 s, ? 1.0054 netbeans x_revs_x000_added_0_copies 6f360122949f 20eb231cc7d0 : 0.000135 s, 0.000133 s, -0.000002 s, ? 0.9852 netbeans x_revs_x_added_x_copies 1ada3faf6fb6 5a39d12eecf4 : 0.000329 s, 0.000320 s, -0.000009 s, ? 0.9726 netbeans x_revs_x00_added_x_copies 35be93ba1e2c 9eec5e90c05f : 0.001343 s, 0.001336 s, -0.000007 s, ? 0.9948 netbeans x000_revs_xx00_added_0_copies eac3045b4fdd 51d4ae7f1290 : 0.029396 s, 0.015573 s, -0.013823 s, ? 0.5298 netbeans x000_revs_x000_added_x_copies e2063d266acd 6081d72689dc : 0.040210 s, 0.018667 s, -0.021543 s, ? 0.4642 netbeans x000_revs_x000_added_x000_copies ff453e9fee32 411350406ec2 : 4.556794 s, 0.112534 s, -4.444260 s, ? 0.0247 netbeans x0000_revs_xx000_added_x000_copies 588c2d1ced70 1aad62e59ddd : killed , 1.231869 s mozilla-central x_revs_x_added_0_copies 3697f962bb7b 7015fcdd43a2 : 0.000199 s, 0.000197 s, -0.000002 s, ? 0.9899 mozilla-central x_revs_x000_added_0_copies dd390860c6c9 40d0c5bed75d : 0.000639 s, 0.000637 s, -0.000002 s, ? 0.9969 mozilla-central x_revs_x_added_x_copies 8d198483ae3b 14207ffc2b2f : 0.000542 s, 0.000303 s, -0.000239 s, ? 0.5590 mozilla-central x_revs_x00_added_x_copies 98cbc58cc6bc 446a150332c3 : 0.001685 s, 0.001663 s, -0.000022 s, ? 0.9869 mozilla-central x_revs_x000_added_x000_copies 3c684b4b8f68 0a5e72d1b479 : 0.006954 s, 0.007008 s, +0.000054 s, ? 1.0078 mozilla-central x_revs_x0000_added_x0000_copies effb563bb7e5 c07a39dc4e80 : 0.132938 s, 0.127385 s, -0.005553 s, ? 0.9582 mozilla-central x000_revs_xx00_added_0_copies 6100d773079a 04a55431795e : 0.008683 s, 0.008740 s, +0.000057 s, ? 1.0066 mozilla-central x000_revs_x000_added_x_copies 9f17a6fc04f9 2d37b966abed : 0.005956 s, 0.005783 s, -0.000173 s, ? 0.9710 mozilla-central x000_revs_x000_added_x000_copies 7c97034feb78 4407bd0c6330 : 0.963905 s, 0.102184 s, -0.861721 s, ? 0.1060 mozilla-central x0000_revs_xx000_added_0_copies 9eec5917337d 67118cc6dcad : 0.049239 s, 0.046220 s, -0.003019 s, ? 0.9387 mozilla-central x0000_revs_xx000_added_x000_copies f78c615a656c 96a38b690156 : 4.217003 s, 0.315271 s, -3.901732 s, ? 0.0748 mozilla-central x00000_revs_x0000_added_x0000_copies 6832ae71433c 4c222a1d9a00 : killed , 3.478747 s mozilla-central x00000_revs_x00000_added_x000_copies 76caed42cf7c 1daa622bbe42 : killed , 4.766435 s mozilla-try x_revs_x_added_0_copies aaf6dde0deb8 9790f499805a : 0.001197 s, 0.001214 s, +0.000017 s, ? 1.0142 mozilla-try x_revs_x000_added_0_copies d8d0222927b4 5bb8ce8c7450 : 0.001213 s, 0.001221 s, +0.000008 s, ? 1.0066 mozilla-try x_revs_x_added_x_copies 092fcca11bdb 936255a0384a : 0.000762 s, 0.000613 s, -0.000149 s, ? 0.8045 mozilla-try x_revs_x00_added_x_copies b53d2fadbdb5 017afae788ec : 0.001909 s, 0.001904 s, -0.000005 s, ? 0.9974 mozilla-try x_revs_x000_added_x000_copies 20408ad61ce5 6f0ee96e21ad : 0.093021 s, 0.093000 s, -0.000021 s, ? 0.9998 mozilla-try x_revs_x0000_added_x0000_copies effb563bb7e5 c07a39dc4e80 : 0.134536 s, 0.132194 s, -0.002342 s, ? 0.9826 mozilla-try x000_revs_xx00_added_0_copies 6100d773079a 04a55431795e : 0.009071 s, 0.009069 s, -0.000002 s, ? 0.9998 mozilla-try x000_revs_x000_added_x_copies 9f17a6fc04f9 2d37b966abed : 0.006206 s, 0.006169 s, -0.000037 s, ? 0.9940 mozilla-try x000_revs_x000_added_x000_copies 1346fd0130e4 4c65cbdabc1f : 1.150502 s, 0.115540 s, -1.034962 s, ? 0.1004 mozilla-try x0000_revs_x_added_0_copies 63519bfd42ee a36a2a865d92 : 1.114864 s, 0.435381 s, -0.679483 s, ? 0.3905 mozilla-try x0000_revs_x_added_x_copies 9fe69ff0762d bcabf2a78927 : 1.042658 s, 0.415461 s, -0.627197 s, ? 0.3985 mozilla-try x0000_revs_xx000_added_x_copies 156f6e2674f2 4d0f2c178e66 : 0.447402 s, 0.155946 s, -0.291456 s, ? 0.3486 mozilla-try x0000_revs_xx000_added_0_copies 9eec5917337d 67118cc6dcad : 0.051132 s, 0.048521 s, -0.002611 s, ? 0.9489 mozilla-try x0000_revs_xx000_added_x000_copies 89294cd501d9 7ccb2fc7ccb5 : 83.508590 s, 9.843481 s, -73.665109 s, ? 0.1179 mozilla-try x0000_revs_x0000_added_x0000_copies e928c65095ed e951f4ad123a : 55.079813 s, 1.465128 s, -53.614685 s, ? 0.0266 mozilla-try x00000_revs_x00000_added_0_copies dc8a3ca7010e d16fde900c9c : 1.442793 s, 1.374283 s, -0.068510 s, ? 0.9525 mozilla-try x00000_revs_x0000_added_x0000_copies 8d3fafa80d4b eb884023b810 : killed , 5.255158 s Repo Cases Source-Rev Dest-Rev Py-time Rust-time Difference Factor ------------------------------------------------------------------------------------------------------------------------------------ mercurial x_revs_x_added_0_copies ad6b123de1c7 39cfcef4f463 : 0.000044 s, 0.000049 s, +0.000005 s, ? 1.1136 mercurial x_revs_x_added_x_copies 2b1c78674230 0c1d10351869 : 0.000138 s, 0.000182 s, +0.000044 s, ? 1.3188 mercurial x000_revs_x000_added_x_copies 81f8ff2a9bf2 dd3267698d84 : 0.005052 s, 0.005872 s, +0.000820 s, ? 1.1623 pypy x_revs_x_added_0_copies aed021ee8ae8 099ed31b181b : 0.000219 s, 0.000229 s, +0.000010 s, ? 1.0457 pypy x_revs_x000_added_0_copies 4aa4e1f8e19a 359343b9ac0e : 0.000055 s, 0.000058 s, +0.000003 s, ? 1.0545 pypy x_revs_x_added_x_copies ac52eb7bbbb0 72e022663155 : 0.000128 s, 0.000148 s, +0.000020 s, ? 1.1562 pypy x_revs_x00_added_x_copies c3b14617fbd7 ace7255d9a26 : 0.001089 s, 0.001205 s, +0.000116 s, ? 1.1065 pypy x_revs_x000_added_x000_copies df6f7a526b60 a83dc6a2d56f : 0.017407 s, 0.025662 s, +0.008255 s, ? 1.4742 pypy x000_revs_xx00_added_0_copies 89a76aede314 2f22446ff07e : 0.094175 s, 0.080113 s, -0.014062 s, ? 0.8507 pypy x000_revs_x000_added_x_copies 8a3b5bfd266e 2c68e87c3efe : 0.238009 s, 0.153030 s, -0.084979 s, ? 0.6430 pypy x000_revs_x000_added_x000_copies 89a76aede314 7b3dda341c84 : 0.125876 s, 0.098774 s, -0.027102 s, ? 0.7847 pypy x0000_revs_x_added_0_copies d1defd0dc478 c9cb1334cc78 : 3.581556 s, 2.780174 s, -0.801382 s, ? 0.7762 pypy x0000_revs_xx000_added_0_copies bf2c629d0071 4ffed77c095c : 0.016721 s, 0.022218 s, +0.005497 s, ? 1.3287 pypy x0000_revs_xx000_added_x000_copies 08ea3258278e d9fa043f30c0 : 0.242367 s, 0.252125 s, +0.009758 s, ? 1.0403 netbeans x_revs_x_added_0_copies fb0955ffcbcd a01e9239f9e7 : 0.000165 s, 0.000186 s, +0.000021 s, ? 1.1273 netbeans x_revs_x000_added_0_copies 6f360122949f 20eb231cc7d0 : 0.000114 s, 0.000133 s, +0.000019 s, ? 1.1667 netbeans x_revs_x_added_x_copies 1ada3faf6fb6 5a39d12eecf4 : 0.000296 s, 0.000320 s, +0.000024 s, ? 1.0811 netbeans x_revs_x00_added_x_copies 35be93ba1e2c 9eec5e90c05f : 0.001124 s, 0.001336 s, +0.000212 s, ? 1.1886 netbeans x000_revs_xx00_added_0_copies eac3045b4fdd 51d4ae7f1290 : 0.013060 s, 0.015573 s, +0.002513 s, ? 1.1924 netbeans x000_revs_x000_added_x_copies e2063d266acd 6081d72689dc : 0.017112 s, 0.018667 s, +0.001555 s, ? 1.0909 netbeans x000_revs_x000_added_x000_copies ff453e9fee32 411350406ec2 : 0.660350 s, 0.112534 s, -0.547816 s, ? 0.1704 netbeans x0000_revs_xx000_added_x000_copies 588c2d1ced70 1aad62e59ddd : 10.032499 s, 1.231869 s, -8.800630 s, ? 0.1228 mozilla-central x_revs_x_added_0_copies 3697f962bb7b 7015fcdd43a2 : 0.000189 s, 0.000197 s, +0.000008 s, ? 1.0423 mozilla-central x_revs_x000_added_0_copies dd390860c6c9 40d0c5bed75d : 0.000462 s, 0.000637 s, +0.000175 s, ? 1.3788 mozilla-central x_revs_x_added_x_copies 8d198483ae3b 14207ffc2b2f : 0.000270 s, 0.000303 s, +0.000033 s, ? 1.1222 mozilla-central x_revs_x00_added_x_copies 98cbc58cc6bc 446a150332c3 : 0.001474 s, 0.001663 s, +0.000189 s, ? 1.1282 mozilla-central x_revs_x000_added_x000_copies 3c684b4b8f68 0a5e72d1b479 : 0.004806 s, 0.007008 s, +0.002202 s, ? 1.4582 mozilla-central x_revs_x0000_added_x0000_copies effb563bb7e5 c07a39dc4e80 : 0.085150 s, 0.127385 s, +0.042235 s, ? 1.4960 mozilla-central x000_revs_xx00_added_0_copies 6100d773079a 04a55431795e : 0.007064 s, 0.008740 s, +0.001676 s, ? 1.2373 mozilla-central x000_revs_x000_added_x_copies 9f17a6fc04f9 2d37b966abed : 0.004741 s, 0.005783 s, +0.001042 s, ? 1.2198 mozilla-central x000_revs_x000_added_x000_copies 7c97034feb78 4407bd0c6330 : 0.190133 s, 0.102184 s, -0.087949 s, ? 0.5374 mozilla-central x0000_revs_xx000_added_0_copies 9eec5917337d 67118cc6dcad : 0.035651 s, 0.046220 s, +0.010569 s, ? 1.2965 mozilla-central x0000_revs_xx000_added_x000_copies f78c615a656c 96a38b690156 : 0.440694 s, 0.315271 s, -0.125423 s, ? 0.7154 mozilla-central x00000_revs_x0000_added_x0000_copies 6832ae71433c 4c222a1d9a00 : 18.454163 s, 3.478747 s, -14.975416 s, ? 0.1885 mozilla-central x00000_revs_x00000_added_x000_copies 76caed42cf7c 1daa622bbe42 : 31.562719 s, 4.766435 s, -26.796284 s, ? 0.1510 mozilla-try x_revs_x_added_0_copies aaf6dde0deb8 9790f499805a : 0.001189 s, 0.001214 s, +0.000025 s, ? 1.0210 mozilla-try x_revs_x000_added_0_copies d8d0222927b4 5bb8ce8c7450 : 0.001204 s, 0.001221 s, +0.000017 s, ? 1.0141 mozilla-try x_revs_x_added_x_copies 092fcca11bdb 936255a0384a : 0.000586 s, 0.000613 s, +0.000027 s, ? 1.0461 mozilla-try x_revs_x00_added_x_copies b53d2fadbdb5 017afae788ec : 0.001845 s, 0.001904 s, +0.000059 s, ? 1.0320 mozilla-try x_revs_x000_added_x000_copies 20408ad61ce5 6f0ee96e21ad : 0.063822 s, 0.093000 s, +0.029178 s, ? 1.4572 mozilla-try x_revs_x0000_added_x0000_copies effb563bb7e5 c07a39dc4e80 : 0.088038 s, 0.132194 s, +0.044156 s, ? 1.5016 mozilla-try x000_revs_xx00_added_0_copies 6100d773079a 04a55431795e : 0.007389 s, 0.009069 s, +0.001680 s, ? 1.2274 mozilla-try x000_revs_x000_added_x_copies 9f17a6fc04f9 2d37b966abed : 0.004868 s, 0.006169 s, +0.001301 s, ? 1.2673 mozilla-try x000_revs_x000_added_x000_copies 1346fd0130e4 4c65cbdabc1f : 0.222450 s, 0.115540 s, -0.106910 s, ? 0.5194 mozilla-try x0000_revs_x_added_0_copies 63519bfd42ee a36a2a865d92 : 0.370675 s, 0.435381 s, +0.064706 s, ? 1.1746 mozilla-try x0000_revs_x_added_x_copies 9fe69ff0762d bcabf2a78927 : 0.358020 s, 0.415461 s, +0.057441 s, ? 1.1604 mozilla-try x0000_revs_xx000_added_x_copies 156f6e2674f2 4d0f2c178e66 : 0.145235 s, 0.155946 s, +0.010711 s, ? 1.0737 mozilla-try x0000_revs_xx000_added_0_copies 9eec5917337d 67118cc6dcad : 0.037606 s, 0.048521 s, +0.010915 s, ? 1.2902 mozilla-try x0000_revs_xx000_added_x000_copies 89294cd501d9 7ccb2fc7ccb5 : 7.382439 s, 9.843481 s, +2.461042 s, ? 1.3334 mozilla-try x0000_revs_x0000_added_x0000_copies e928c65095ed e951f4ad123a : 7.273506 s, 1.465128 s, -5.808378 s, ? 0.2014 mozilla-try x00000_revs_x00000_added_0_copies dc8a3ca7010e d16fde900c9c : 1.074593 s, 1.374283 s, +0.299690 s, ? 1.2789 mozilla-try x00000_revs_x0000_added_x0000_copies 8d3fafa80d4b eb884023b810 : 27.746195 s, 5.255158 s, -22.491037 s, ? 0.1894 Repo Case Source-Rev Dest-Rev filelog sidedata Difference Factor --------------------------------------------------------------------------------------------------------------------------------------- mercurial x_revs_x_added_0_copies ad6b123de1c7 39cfcef4f463 : 0.000906 s, 0.000049 s, -0.000857 s, ? 0.054084 mercurial x_revs_x_added_x_copies 2b1c78674230 0c1d10351869 : 0.001832 s, 0.000182 s, -0.001650 s, ? 0.099345 mercurial x000_revs_x000_added_x_copies 81f8ff2a9bf2 dd3267698d84 : 0.018101 s, 0.005872 s, -0.012229 s, ? 0.324402 pypy x_revs_x_added_0_copies aed021ee8ae8 099ed31b181b : 0.001508 s, 0.000229 s, -0.001279 s, ? 0.151857 pypy x_revs_x000_added_0_copies 4aa4e1f8e19a 359343b9ac0e : 0.207438 s, 0.000058 s, -0.207380 s, ? 0.000280 pypy x_revs_x_added_x_copies ac52eb7bbbb0 72e022663155 : 0.016956 s, 0.000148 s, -0.016808 s, ? 0.008728 pypy x_revs_x00_added_x_copies c3b14617fbd7 ace7255d9a26 : 0.019129 s, 0.001205 s, -0.017924 s, ? 0.062993 pypy x_revs_x000_added_x000_copies df6f7a526b60 a83dc6a2d56f : 0.772050 s, 0.025662 s, -0.746388 s, ? 0.033239 pypy x000_revs_xx00_added_0_copies 89a76aede314 2f22446ff07e : 1.181775 s, 0.080113 s, -1.101662 s, ? 0.067790 pypy x000_revs_x000_added_x_copies 8a3b5bfd266e 2c68e87c3efe : 1.255717 s, 0.153030 s, -1.102687 s, ? 0.121867 pypy x000_revs_x000_added_x000_copies 89a76aede314 7b3dda341c84 : 1.628132 s, 0.098774 s, -1.529358 s, ? 0.060667 pypy x0000_revs_x_added_0_copies d1defd0dc478 c9cb1334cc78 : 0.001055 s, 2.780174 s, +2.779119 s, ? 2635.236019 pypy x0000_revs_xx000_added_0_copies bf2c629d0071 4ffed77c095c : 1.056756 s, 0.022218 s, -1.034538 s, ? 0.021025 pypy x0000_revs_xx000_added_x000_copies 08ea3258278e d9fa043f30c0 : 1.354211 s, 0.252125 s, -1.102086 s, ? 0.186179 netbeans x_revs_x_added_0_copies fb0955ffcbcd a01e9239f9e7 : 0.027759 s, 0.000186 s, -0.027573 s, ? 0.006701 netbeans x_revs_x000_added_0_copies 6f360122949f 20eb231cc7d0 : 0.133382 s, 0.000133 s, -0.133249 s, ? 0.000997 netbeans x_revs_x_added_x_copies 1ada3faf6fb6 5a39d12eecf4 : 0.025353 s, 0.000320 s, -0.025033 s, ? 0.012622 netbeans x_revs_x00_added_x_copies 35be93ba1e2c 9eec5e90c05f : 0.052939 s, 0.001336 s, -0.051603 s, ? 0.025237 netbeans x000_revs_xx00_added_0_copies eac3045b4fdd 51d4ae7f1290 : 0.037532 s, 0.015573 s, -0.021959 s, ? 0.414926 netbeans x000_revs_x000_added_x_copies e2063d266acd 6081d72689dc : 0.195401 s, 0.018667 s, -0.176734 s, ? 0.095532 netbeans x000_revs_x000_added_x000_copies ff453e9fee32 411350406ec2 : 0.939593 s, 0.112534 s, -0.827059 s, ? 0.119769 netbeans x0000_revs_xx000_added_x000_copies 588c2d1ced70 1aad62e59ddd : 3.824967 s, 1.231869 s, -2.593098 s, ? 0.322060 mozilla-central x_revs_x_added_0_copies 3697f962bb7b 7015fcdd43a2 : 0.024626 s, 0.000197 s, -0.024429 s, ? 0.008000 mozilla-central x_revs_x000_added_0_copies dd390860c6c9 40d0c5bed75d : 0.141098 s, 0.000637 s, -0.140461 s, ? 0.004515 mozilla-central x_revs_x_added_x_copies 8d198483ae3b 14207ffc2b2f : 0.025973 s, 0.000303 s, -0.025670 s, ? 0.011666 mozilla-central x_revs_x00_added_x_copies 98cbc58cc6bc 446a150332c3 : 0.085098 s, 0.001663 s, -0.083435 s, ? 0.019542 mozilla-central x_revs_x000_added_x000_copies 3c684b4b8f68 0a5e72d1b479 : 0.196623 s, 0.007008 s, -0.189615 s, ? 0.035642 mozilla-central x_revs_x0000_added_x0000_copies effb563bb7e5 c07a39dc4e80 : 2.182911 s, 0.127385 s, -2.055526 s, ? 0.058356 mozilla-central x000_revs_xx00_added_0_copies 6100d773079a 04a55431795e : 0.088865 s, 0.008740 s, -0.080125 s, ? 0.098351 mozilla-central x000_revs_x000_added_x_copies 9f17a6fc04f9 2d37b966abed : 0.731211 s, 0.005783 s, -0.725428 s, ? 0.007909 mozilla-central x000_revs_x000_added_x000_copies 7c97034feb78 4407bd0c6330 : 1.142414 s, 0.102184 s, -1.040230 s, ? 0.089446 mozilla-central x0000_revs_xx000_added_0_copies 9eec5917337d 67118cc6dcad : 6.667504 s, 0.046220 s, -6.621284 s, ? 0.006932 mozilla-central x0000_revs_xx000_added_x000_copies f78c615a656c 96a38b690156 : 3.267274 s, 0.315271 s, -2.952003 s, ? 0.096494 mozilla-central x00000_revs_x0000_added_x0000_copies 6832ae71433c 4c222a1d9a00 : 16.038104 s, 3.478747 s, -12.559357 s, ? 0.216905 mozilla-central x00000_revs_x00000_added_x000_copies 76caed42cf7c 1daa622bbe42 : 20.639928 s, 4.766435 s, -15.873493 s, ? 0.230933 mozilla-try x_revs_x_added_0_copies aaf6dde0deb8 9790f499805a : 0.080513 s, 0.001214 s, -0.079299 s, ? 0.015078 mozilla-try x_revs_x000_added_0_copies d8d0222927b4 5bb8ce8c7450 : 0.502965 s, 0.001221 s, -0.501744 s, ? 0.002428 mozilla-try x_revs_x_added_x_copies 092fcca11bdb 936255a0384a : 0.021408 s, 0.000613 s, -0.020795 s, ? 0.028634 mozilla-try x_revs_x00_added_x_copies b53d2fadbdb5 017afae788ec : 0.228498 s, 0.001904 s, -0.226594 s, ? 0.008333 mozilla-try x_revs_x000_added_x000_copies 20408ad61ce5 6f0ee96e21ad : 1.118490 s, 0.093000 s, -1.025490 s, ? 0.083148 mozilla-try x_revs_x0000_added_x0000_copies effb563bb7e5 c07a39dc4e80 : 2.188390 s, 0.132194 s, -2.056196 s, ? 0.060407 mozilla-try x000_revs_xx00_added_0_copies 6100d773079a 04a55431795e : 0.089755 s, 0.009069 s, -0.080686 s, ? 0.101042 mozilla-try x000_revs_x000_added_x_copies 9f17a6fc04f9 2d37b966abed : 0.750223 s, 0.006169 s, -0.744054 s, ? 0.008223 mozilla-try x000_revs_x000_added_x000_copies 1346fd0130e4 4c65cbdabc1f : 1.189708 s, 0.115540 s, -1.074168 s, ? 0.097116 mozilla-try x0000_revs_x_added_0_copies 63519bfd42ee a36a2a865d92 : 0.088715 s, 0.435381 s, +0.346666 s, ? 4.907637 mozilla-try x0000_revs_x_added_x_copies 9fe69ff0762d bcabf2a78927 : 0.080765 s, 0.415461 s, +0.334696 s, ? 5.144072 mozilla-try x0000_revs_xx000_added_x_copies 156f6e2674f2 4d0f2c178e66 : 7.487310 s, 0.155946 s, -7.331364 s, ? 0.020828 mozilla-try x0000_revs_xx000_added_0_copies 9eec5917337d 67118cc6dcad : 6.676682 s, 0.048521 s, -6.628161 s, ? 0.007267 mozilla-try x0000_revs_xx000_added_x000_copies 89294cd501d9 7ccb2fc7ccb5 : 7.598509 s, 9.843481 s, +2.244972 s, ? 1.295449 mozilla-try x0000_revs_x0000_added_x0000_copies e928c65095ed e951f4ad123a : 9.871850 s, 1.465128 s, -8.406722 s, ? 0.148415 mozilla-try x00000_revs_x_added_0_copies 6a320851d377 1ebb79acd503 : 0.092232 s, killed mozilla-try x00000_revs_x00000_added_0_copies dc8a3ca7010e d16fde900c9c : 26.471560 s, 1.374283 s, -25.097277 s, ? 0.051915 mozilla-try x00000_revs_x_added_x_copies 5173c4b6f97c 95d83ee7242d : 0.093892 s, killed mozilla-try x00000_revs_x000_added_x_copies 9126823d0e9c ca82787bb23c : 0.227503 s, killed mozilla-try x00000_revs_x0000_added_x0000_copies 8d3fafa80d4b eb884023b810 : 19.064699 s, 5.255158 s, -13.809541 s, ? 0.275649 mozilla-try x00000_revs_x00000_added_x0000_copies 1b661134e2ca 1ae03d022d6d : 21.145391 s, killed mozilla-try x00000_revs_x00000_added_x000_copies 9b2a99adc05e 8e29777b48e6 : 25.304164 s, killed Differential Revision: https://phab.mercurial-scm.org/D9302
author Pierre-Yves David <pierre-yves.david@octobus.net>
date Tue, 21 Apr 2020 15:00:32 +0200
parents 2d5dfc8fed55
children 2e2033081274
line wrap: on
line source

// hg_path.rs
//
// Copyright 2019 Raphaël Gomès <rgomes@octobus.net>
//
// This software may be used and distributed according to the terms of the
// GNU General Public License version 2 or any later version.

use std::borrow::Borrow;
use std::convert::TryFrom;
use std::ffi::{OsStr, OsString};
use std::fmt;
use std::ops::Deref;
use std::path::{Path, PathBuf};

#[derive(Debug, Eq, PartialEq)]
pub enum HgPathError {
    /// Bytes from the invalid `HgPath`
    LeadingSlash(Vec<u8>),
    ConsecutiveSlashes {
        bytes: Vec<u8>,
        second_slash_index: usize,
    },
    ContainsNullByte {
        bytes: Vec<u8>,
        null_byte_index: usize,
    },
    /// Bytes
    DecodeError(Vec<u8>),
    /// The rest come from audit errors
    EndsWithSlash(HgPathBuf),
    ContainsIllegalComponent(HgPathBuf),
    /// Path is inside the `.hg` folder
    InsideDotHg(HgPathBuf),
    IsInsideNestedRepo {
        path: HgPathBuf,
        nested_repo: HgPathBuf,
    },
    TraversesSymbolicLink {
        path: HgPathBuf,
        symlink: HgPathBuf,
    },
    NotFsCompliant(HgPathBuf),
    /// `path` is the smallest invalid path
    NotUnderRoot {
        path: PathBuf,
        root: PathBuf,
    },
}

impl ToString for HgPathError {
    fn to_string(&self) -> String {
        match self {
            HgPathError::LeadingSlash(bytes) => {
                format!("Invalid HgPath '{:?}': has a leading slash.", bytes)
            }
            HgPathError::ConsecutiveSlashes {
                bytes,
                second_slash_index: pos,
            } => format!(
                "Invalid HgPath '{:?}': consecutive slashes at pos {}.",
                bytes, pos
            ),
            HgPathError::ContainsNullByte {
                bytes,
                null_byte_index: pos,
            } => format!(
                "Invalid HgPath '{:?}': contains null byte at pos {}.",
                bytes, pos
            ),
            HgPathError::DecodeError(bytes) => {
                format!("Invalid HgPath '{:?}': could not be decoded.", bytes)
            }
            HgPathError::EndsWithSlash(path) => {
                format!("Audit failed for '{}': ends with a slash.", path)
            }
            HgPathError::ContainsIllegalComponent(path) => format!(
                "Audit failed for '{}': contains an illegal component.",
                path
            ),
            HgPathError::InsideDotHg(path) => format!(
                "Audit failed for '{}': is inside the '.hg' folder.",
                path
            ),
            HgPathError::IsInsideNestedRepo {
                path,
                nested_repo: nested,
            } => format!(
                "Audit failed for '{}': is inside a nested repository '{}'.",
                path, nested
            ),
            HgPathError::TraversesSymbolicLink { path, symlink } => format!(
                "Audit failed for '{}': traverses symbolic link '{}'.",
                path, symlink
            ),
            HgPathError::NotFsCompliant(path) => format!(
                "Audit failed for '{}': cannot be turned into a \
                 filesystem path.",
                path
            ),
            HgPathError::NotUnderRoot { path, root } => format!(
                "Audit failed for '{}': not under root {}.",
                path.display(),
                root.display()
            ),
        }
    }
}

impl From<HgPathError> for std::io::Error {
    fn from(e: HgPathError) -> Self {
        std::io::Error::new(std::io::ErrorKind::InvalidData, e.to_string())
    }
}

/// This is a repository-relative path (or canonical path):
///     - no null characters
///     - `/` separates directories
///     - no consecutive slashes
///     - no leading slash,
///     - no `.` nor `..` of special meaning
///     - stored in repository and shared across platforms
///
/// Note: there is no guarantee of any `HgPath` being well-formed at any point
/// in its lifetime for performance reasons and to ease ergonomics. It is
/// however checked using the `check_state` method before any file-system
/// operation.
///
/// This allows us to be encoding-transparent as much as possible, until really
/// needed; `HgPath` can be transformed into a platform-specific path (`OsStr`
/// or `Path`) whenever more complex operations are needed:
/// On Unix, it's just byte-to-byte conversion. On Windows, it has to be
/// decoded from MBCS to WTF-8. If WindowsUTF8Plan is implemented, the source
/// character encoding will be determined on a per-repository basis.
//
// FIXME: (adapted from a comment in the stdlib)
// `HgPath::new()` current implementation relies on `Slice` being
// layout-compatible with `[u8]`.
// When attribute privacy is implemented, `Slice` should be annotated as
// `#[repr(transparent)]`.
// Anyway, `Slice` representation and layout are considered implementation
// detail, are not documented and must not be relied upon.
#[derive(Eq, Ord, PartialEq, PartialOrd, Hash)]
pub struct HgPath {
    inner: [u8],
}

impl HgPath {
    pub fn new<S: AsRef<[u8]> + ?Sized>(s: &S) -> &Self {
        unsafe { &*(s.as_ref() as *const [u8] as *const Self) }
    }
    pub fn is_empty(&self) -> bool {
        self.inner.is_empty()
    }
    pub fn len(&self) -> usize {
        self.inner.len()
    }
    fn to_hg_path_buf(&self) -> HgPathBuf {
        HgPathBuf {
            inner: self.inner.to_owned(),
        }
    }
    pub fn bytes(&self) -> std::slice::Iter<u8> {
        self.inner.iter()
    }
    pub fn to_ascii_uppercase(&self) -> HgPathBuf {
        HgPathBuf::from(self.inner.to_ascii_uppercase())
    }
    pub fn to_ascii_lowercase(&self) -> HgPathBuf {
        HgPathBuf::from(self.inner.to_ascii_lowercase())
    }
    pub fn as_bytes(&self) -> &[u8] {
        &self.inner
    }
    pub fn contains(&self, other: u8) -> bool {
        self.inner.contains(&other)
    }
    pub fn starts_with(&self, needle: impl AsRef<Self>) -> bool {
        self.inner.starts_with(needle.as_ref().as_bytes())
    }
    pub fn trim_trailing_slash(&self) -> &Self {
        Self::new(if self.inner.last() == Some(&b'/') {
            &self.inner[..self.inner.len() - 1]
        } else {
            &self.inner[..]
        })
    }
    /// Returns a tuple of slices `(base, filename)` resulting from the split
    /// at the rightmost `/`, if any.
    ///
    /// # Examples:
    ///
    /// ```
    /// use hg::utils::hg_path::HgPath;
    ///
    /// let path = HgPath::new(b"cool/hg/path").split_filename();
    /// assert_eq!(path, (HgPath::new(b"cool/hg"), HgPath::new(b"path")));
    ///
    /// let path = HgPath::new(b"pathwithoutsep").split_filename();
    /// assert_eq!(path, (HgPath::new(b""), HgPath::new(b"pathwithoutsep")));
    /// ```
    pub fn split_filename(&self) -> (&Self, &Self) {
        match &self.inner.iter().rposition(|c| *c == b'/') {
            None => (HgPath::new(""), &self),
            Some(size) => (
                HgPath::new(&self.inner[..*size]),
                HgPath::new(&self.inner[*size + 1..]),
            ),
        }
    }
    pub fn join<T: ?Sized + AsRef<Self>>(&self, other: &T) -> HgPathBuf {
        let mut inner = self.inner.to_owned();
        if !inner.is_empty() && inner.last() != Some(&b'/') {
            inner.push(b'/');
        }
        inner.extend(other.as_ref().bytes());
        HgPathBuf::from_bytes(&inner)
    }
    pub fn parent(&self) -> &Self {
        let inner = self.as_bytes();
        HgPath::new(match inner.iter().rposition(|b| *b == b'/') {
            Some(pos) => &inner[..pos],
            None => &[],
        })
    }
    /// Given a base directory, returns the slice of `self` relative to the
    /// base directory. If `base` is not a directory (does not end with a
    /// `b'/'`), returns `None`.
    pub fn relative_to(&self, base: impl AsRef<Self>) -> Option<&Self> {
        let base = base.as_ref();
        if base.is_empty() {
            return Some(self);
        }
        let is_dir = base.as_bytes().ends_with(b"/");
        if is_dir && self.starts_with(base) {
            Some(Self::new(&self.inner[base.len()..]))
        } else {
            None
        }
    }

    #[cfg(windows)]
    /// Copied from the Python stdlib's `os.path.splitdrive` implementation.
    ///
    /// Split a pathname into drive/UNC sharepoint and relative path
    /// specifiers. Returns a 2-tuple (drive_or_unc, path); either part may
    /// be empty.
    ///
    /// If you assign
    ///  result = split_drive(p)
    /// It is always true that:
    ///  result[0] + result[1] == p
    ///
    /// If the path contained a drive letter, drive_or_unc will contain
    /// everything up to and including the colon.
    /// e.g. split_drive("c:/dir") returns ("c:", "/dir")
    ///
    /// If the path contained a UNC path, the drive_or_unc will contain the
    /// host name and share up to but not including the fourth directory
    /// separator character.
    /// e.g. split_drive("//host/computer/dir") returns ("//host/computer",
    /// "/dir")
    ///
    /// Paths cannot contain both a drive letter and a UNC path.
    pub fn split_drive<'a>(&self) -> (&HgPath, &HgPath) {
        let bytes = self.as_bytes();
        let is_sep = |b| std::path::is_separator(b as char);

        if self.len() < 2 {
            (HgPath::new(b""), &self)
        } else if is_sep(bytes[0])
            && is_sep(bytes[1])
            && (self.len() == 2 || !is_sep(bytes[2]))
        {
            // Is a UNC path:
            // vvvvvvvvvvvvvvvvvvvv drive letter or UNC path
            // \\machine\mountpoint\directory\etc\...
            //           directory ^^^^^^^^^^^^^^^

            let machine_end_index = bytes[2..].iter().position(|b| is_sep(*b));
            let mountpoint_start_index = if let Some(i) = machine_end_index {
                i + 2
            } else {
                return (HgPath::new(b""), &self);
            };

            match bytes[mountpoint_start_index + 1..]
                .iter()
                .position(|b| is_sep(*b))
            {
                // A UNC path can't have two slashes in a row
                // (after the initial two)
                Some(0) => (HgPath::new(b""), &self),
                Some(i) => {
                    let (a, b) =
                        bytes.split_at(mountpoint_start_index + 1 + i);
                    (HgPath::new(a), HgPath::new(b))
                }
                None => (&self, HgPath::new(b"")),
            }
        } else if bytes[1] == b':' {
            // Drive path c:\directory
            let (a, b) = bytes.split_at(2);
            (HgPath::new(a), HgPath::new(b))
        } else {
            (HgPath::new(b""), &self)
        }
    }

    #[cfg(unix)]
    /// Split a pathname into drive and path. On Posix, drive is always empty.
    pub fn split_drive(&self) -> (&HgPath, &HgPath) {
        (HgPath::new(b""), &self)
    }

    /// Checks for errors in the path, short-circuiting at the first one.
    /// This generates fine-grained errors useful for debugging.
    /// To simply check if the path is valid during tests, use `is_valid`.
    pub fn check_state(&self) -> Result<(), HgPathError> {
        if self.is_empty() {
            return Ok(());
        }
        let bytes = self.as_bytes();
        let mut previous_byte = None;

        if bytes[0] == b'/' {
            return Err(HgPathError::LeadingSlash(bytes.to_vec()));
        }
        for (index, byte) in bytes.iter().enumerate() {
            match byte {
                0 => {
                    return Err(HgPathError::ContainsNullByte {
                        bytes: bytes.to_vec(),
                        null_byte_index: index,
                    })
                }
                b'/' => {
                    if previous_byte.is_some() && previous_byte == Some(b'/') {
                        return Err(HgPathError::ConsecutiveSlashes {
                            bytes: bytes.to_vec(),
                            second_slash_index: index,
                        });
                    }
                }
                _ => (),
            };
            previous_byte = Some(*byte);
        }
        Ok(())
    }

    #[cfg(test)]
    /// Only usable during tests to force developers to handle invalid states
    fn is_valid(&self) -> bool {
        self.check_state().is_ok()
    }
}

impl fmt::Debug for HgPath {
    fn fmt(&self, f: &mut fmt::Formatter<'_>) -> fmt::Result {
        write!(f, "HgPath({:?})", String::from_utf8_lossy(&self.inner))
    }
}

impl fmt::Display for HgPath {
    fn fmt(&self, f: &mut fmt::Formatter<'_>) -> fmt::Result {
        write!(f, "{}", String::from_utf8_lossy(&self.inner))
    }
}

#[derive(Default, Eq, Ord, Clone, PartialEq, PartialOrd, Hash)]
pub struct HgPathBuf {
    inner: Vec<u8>,
}

impl HgPathBuf {
    pub fn new() -> Self {
        Default::default()
    }
    pub fn push(&mut self, byte: u8) {
        self.inner.push(byte);
    }
    pub fn from_bytes(s: &[u8]) -> HgPathBuf {
        HgPath::new(s).to_owned()
    }
    pub fn into_vec(self) -> Vec<u8> {
        self.inner
    }
}

impl fmt::Debug for HgPathBuf {
    fn fmt(&self, f: &mut fmt::Formatter<'_>) -> fmt::Result {
        write!(f, "HgPathBuf({:?})", String::from_utf8_lossy(&self.inner))
    }
}

impl fmt::Display for HgPathBuf {
    fn fmt(&self, f: &mut fmt::Formatter<'_>) -> fmt::Result {
        write!(f, "{}", String::from_utf8_lossy(&self.inner))
    }
}

impl Deref for HgPathBuf {
    type Target = HgPath;

    #[inline]
    fn deref(&self) -> &HgPath {
        &HgPath::new(&self.inner)
    }
}

impl From<Vec<u8>> for HgPathBuf {
    fn from(vec: Vec<u8>) -> Self {
        Self { inner: vec }
    }
}

impl<T: ?Sized + AsRef<HgPath>> From<&T> for HgPathBuf {
    fn from(s: &T) -> HgPathBuf {
        s.as_ref().to_owned()
    }
}

impl Into<Vec<u8>> for HgPathBuf {
    fn into(self) -> Vec<u8> {
        self.inner
    }
}

impl Borrow<HgPath> for HgPathBuf {
    fn borrow(&self) -> &HgPath {
        &HgPath::new(self.as_bytes())
    }
}

impl ToOwned for HgPath {
    type Owned = HgPathBuf;

    fn to_owned(&self) -> HgPathBuf {
        self.to_hg_path_buf()
    }
}

impl AsRef<HgPath> for HgPath {
    fn as_ref(&self) -> &HgPath {
        self
    }
}

impl AsRef<HgPath> for HgPathBuf {
    fn as_ref(&self) -> &HgPath {
        self
    }
}

impl Extend<u8> for HgPathBuf {
    fn extend<T: IntoIterator<Item = u8>>(&mut self, iter: T) {
        self.inner.extend(iter);
    }
}

/// TODO: Once https://www.mercurial-scm.org/wiki/WindowsUTF8Plan is
/// implemented, these conversion utils will have to work differently depending
/// on the repository encoding: either `UTF-8` or `MBCS`.

pub fn hg_path_to_os_string<P: AsRef<HgPath>>(
    hg_path: P,
) -> Result<OsString, HgPathError> {
    hg_path.as_ref().check_state()?;
    let os_str;
    #[cfg(unix)]
    {
        use std::os::unix::ffi::OsStrExt;
        os_str = std::ffi::OsStr::from_bytes(&hg_path.as_ref().as_bytes());
    }
    // TODO Handle other platforms
    // TODO: convert from WTF8 to Windows MBCS (ANSI encoding).
    Ok(os_str.to_os_string())
}

pub fn hg_path_to_path_buf<P: AsRef<HgPath>>(
    hg_path: P,
) -> Result<PathBuf, HgPathError> {
    Ok(Path::new(&hg_path_to_os_string(hg_path)?).to_path_buf())
}

pub fn os_string_to_hg_path_buf<S: AsRef<OsStr>>(
    os_string: S,
) -> Result<HgPathBuf, HgPathError> {
    let buf;
    #[cfg(unix)]
    {
        use std::os::unix::ffi::OsStrExt;
        buf = HgPathBuf::from_bytes(&os_string.as_ref().as_bytes());
    }
    // TODO Handle other platforms
    // TODO: convert from WTF8 to Windows MBCS (ANSI encoding).

    buf.check_state()?;
    Ok(buf)
}

pub fn path_to_hg_path_buf<P: AsRef<Path>>(
    path: P,
) -> Result<HgPathBuf, HgPathError> {
    let buf;
    let os_str = path.as_ref().as_os_str();
    #[cfg(unix)]
    {
        use std::os::unix::ffi::OsStrExt;
        buf = HgPathBuf::from_bytes(&os_str.as_bytes());
    }
    // TODO Handle other platforms
    // TODO: convert from WTF8 to Windows MBCS (ANSI encoding).

    buf.check_state()?;
    Ok(buf)
}

impl TryFrom<PathBuf> for HgPathBuf {
    type Error = HgPathError;
    fn try_from(path: PathBuf) -> Result<Self, Self::Error> {
        path_to_hg_path_buf(path)
    }
}

#[cfg(test)]
mod tests {
    use super::*;
    use pretty_assertions::assert_eq;

    #[test]
    fn test_path_states() {
        assert_eq!(
            Err(HgPathError::LeadingSlash(b"/".to_vec())),
            HgPath::new(b"/").check_state()
        );
        assert_eq!(
            Err(HgPathError::ConsecutiveSlashes {
                bytes: b"a/b//c".to_vec(),
                second_slash_index: 4
            }),
            HgPath::new(b"a/b//c").check_state()
        );
        assert_eq!(
            Err(HgPathError::ContainsNullByte {
                bytes: b"a/b/\0c".to_vec(),
                null_byte_index: 4
            }),
            HgPath::new(b"a/b/\0c").check_state()
        );
        // TODO test HgPathError::DecodeError for the Windows implementation.
        assert_eq!(true, HgPath::new(b"").is_valid());
        assert_eq!(true, HgPath::new(b"a/b/c").is_valid());
        // Backslashes in paths are not significant, but allowed
        assert_eq!(true, HgPath::new(br"a\b/c").is_valid());
        // Dots in paths are not significant, but allowed
        assert_eq!(true, HgPath::new(b"a/b/../c/").is_valid());
        assert_eq!(true, HgPath::new(b"./a/b/../c/").is_valid());
    }

    #[test]
    fn test_iter() {
        let path = HgPath::new(b"a");
        let mut iter = path.bytes();
        assert_eq!(Some(&b'a'), iter.next());
        assert_eq!(None, iter.next_back());
        assert_eq!(None, iter.next());

        let path = HgPath::new(b"a");
        let mut iter = path.bytes();
        assert_eq!(Some(&b'a'), iter.next_back());
        assert_eq!(None, iter.next_back());
        assert_eq!(None, iter.next());

        let path = HgPath::new(b"abc");
        let mut iter = path.bytes();
        assert_eq!(Some(&b'a'), iter.next());
        assert_eq!(Some(&b'c'), iter.next_back());
        assert_eq!(Some(&b'b'), iter.next_back());
        assert_eq!(None, iter.next_back());
        assert_eq!(None, iter.next());

        let path = HgPath::new(b"abc");
        let mut iter = path.bytes();
        assert_eq!(Some(&b'a'), iter.next());
        assert_eq!(Some(&b'b'), iter.next());
        assert_eq!(Some(&b'c'), iter.next());
        assert_eq!(None, iter.next_back());
        assert_eq!(None, iter.next());

        let path = HgPath::new(b"abc");
        let iter = path.bytes();
        let mut vec = Vec::new();
        vec.extend(iter);
        assert_eq!(vec![b'a', b'b', b'c'], vec);

        let path = HgPath::new(b"abc");
        let mut iter = path.bytes();
        assert_eq!(Some(2), iter.rposition(|c| *c == b'c'));

        let path = HgPath::new(b"abc");
        let mut iter = path.bytes();
        assert_eq!(None, iter.rposition(|c| *c == b'd'));
    }

    #[test]
    fn test_join() {
        let path = HgPathBuf::from_bytes(b"a").join(HgPath::new(b"b"));
        assert_eq!(b"a/b", path.as_bytes());

        let path = HgPathBuf::from_bytes(b"a/").join(HgPath::new(b"b/c"));
        assert_eq!(b"a/b/c", path.as_bytes());

        // No leading slash if empty before join
        let path = HgPathBuf::new().join(HgPath::new(b"b/c"));
        assert_eq!(b"b/c", path.as_bytes());

        // The leading slash is an invalid representation of an `HgPath`, but
        // it can happen. This creates another invalid representation of
        // consecutive bytes.
        // TODO What should be done in this case? Should we silently remove
        // the extra slash? Should we change the signature to a problematic
        // `Result<HgPathBuf, HgPathError>`, or should we just keep it so and
        // let the error happen upon filesystem interaction?
        let path = HgPathBuf::from_bytes(b"a/").join(HgPath::new(b"/b"));
        assert_eq!(b"a//b", path.as_bytes());
        let path = HgPathBuf::from_bytes(b"a").join(HgPath::new(b"/b"));
        assert_eq!(b"a//b", path.as_bytes());
    }

    #[test]
    fn test_relative_to() {
        let path = HgPath::new(b"");
        let base = HgPath::new(b"");
        assert_eq!(Some(path), path.relative_to(base));

        let path = HgPath::new(b"path");
        let base = HgPath::new(b"");
        assert_eq!(Some(path), path.relative_to(base));

        let path = HgPath::new(b"a");
        let base = HgPath::new(b"b");
        assert_eq!(None, path.relative_to(base));

        let path = HgPath::new(b"a/b");
        let base = HgPath::new(b"a");
        assert_eq!(None, path.relative_to(base));

        let path = HgPath::new(b"a/b");
        let base = HgPath::new(b"a/");
        assert_eq!(Some(HgPath::new(b"b")), path.relative_to(base));

        let path = HgPath::new(b"nested/path/to/b");
        let base = HgPath::new(b"nested/path/");
        assert_eq!(Some(HgPath::new(b"to/b")), path.relative_to(base));

        let path = HgPath::new(b"ends/with/dir/");
        let base = HgPath::new(b"ends/");
        assert_eq!(Some(HgPath::new(b"with/dir/")), path.relative_to(base));
    }

    #[test]
    #[cfg(unix)]
    fn test_split_drive() {
        // Taken from the Python stdlib's tests
        assert_eq!(
            HgPath::new(br"/foo/bar").split_drive(),
            (HgPath::new(b""), HgPath::new(br"/foo/bar"))
        );
        assert_eq!(
            HgPath::new(br"foo:bar").split_drive(),
            (HgPath::new(b""), HgPath::new(br"foo:bar"))
        );
        assert_eq!(
            HgPath::new(br":foo:bar").split_drive(),
            (HgPath::new(b""), HgPath::new(br":foo:bar"))
        );
        // Also try NT paths; should not split them
        assert_eq!(
            HgPath::new(br"c:\foo\bar").split_drive(),
            (HgPath::new(b""), HgPath::new(br"c:\foo\bar"))
        );
        assert_eq!(
            HgPath::new(b"c:/foo/bar").split_drive(),
            (HgPath::new(b""), HgPath::new(br"c:/foo/bar"))
        );
        assert_eq!(
            HgPath::new(br"\\conky\mountpoint\foo\bar").split_drive(),
            (
                HgPath::new(b""),
                HgPath::new(br"\\conky\mountpoint\foo\bar")
            )
        );
    }

    #[test]
    #[cfg(windows)]
    fn test_split_drive() {
        assert_eq!(
            HgPath::new(br"c:\foo\bar").split_drive(),
            (HgPath::new(br"c:"), HgPath::new(br"\foo\bar"))
        );
        assert_eq!(
            HgPath::new(b"c:/foo/bar").split_drive(),
            (HgPath::new(br"c:"), HgPath::new(br"/foo/bar"))
        );
        assert_eq!(
            HgPath::new(br"\\conky\mountpoint\foo\bar").split_drive(),
            (
                HgPath::new(br"\\conky\mountpoint"),
                HgPath::new(br"\foo\bar")
            )
        );
        assert_eq!(
            HgPath::new(br"//conky/mountpoint/foo/bar").split_drive(),
            (
                HgPath::new(br"//conky/mountpoint"),
                HgPath::new(br"/foo/bar")
            )
        );
        assert_eq!(
            HgPath::new(br"\\\conky\mountpoint\foo\bar").split_drive(),
            (
                HgPath::new(br""),
                HgPath::new(br"\\\conky\mountpoint\foo\bar")
            )
        );
        assert_eq!(
            HgPath::new(br"///conky/mountpoint/foo/bar").split_drive(),
            (
                HgPath::new(br""),
                HgPath::new(br"///conky/mountpoint/foo/bar")
            )
        );
        assert_eq!(
            HgPath::new(br"\\conky\\mountpoint\foo\bar").split_drive(),
            (
                HgPath::new(br""),
                HgPath::new(br"\\conky\\mountpoint\foo\bar")
            )
        );
        assert_eq!(
            HgPath::new(br"//conky//mountpoint/foo/bar").split_drive(),
            (
                HgPath::new(br""),
                HgPath::new(br"//conky//mountpoint/foo/bar")
            )
        );
        // UNC part containing U+0130
        assert_eq!(
            HgPath::new(b"//conky/MOUNTPO\xc4\xb0NT/foo/bar").split_drive(),
            (
                HgPath::new(b"//conky/MOUNTPO\xc4\xb0NT"),
                HgPath::new(br"/foo/bar")
            )
        );
    }

    #[test]
    fn test_parent() {
        let path = HgPath::new(b"");
        assert_eq!(path.parent(), path);

        let path = HgPath::new(b"a");
        assert_eq!(path.parent(), HgPath::new(b""));

        let path = HgPath::new(b"a/b");
        assert_eq!(path.parent(), HgPath::new(b"a"));

        let path = HgPath::new(b"a/other/b");
        assert_eq!(path.parent(), HgPath::new(b"a/other"));
    }
}