Repository navigation
fix #2762 - #2767
fix #2762#2767
Conversation
|
Detailed content: #2762 |
|
@appleboy @rw-access Would you please take a look when you are free? Very thx. 😃 |
rw-access
left a comment
There was a problem hiding this comment.
can you add some more comments and/or an explanation of how this works? it's not clear what's going on and that makes it hard to review
|
Please also add benchmark result like #2706 (comment) comment. |
hi,local 'go test --run TestTreeWildcard/TestRunDynamicRouting' passed and add benchmark if there are omissions, please help me to add. 😆 The previous code could not match when running 'benchmark', and 404 not found occurred,Previously, the matching of 'level 1 router' was unreliable, so the matching method was changed😆 The benchmarks: this PR - https://travis-ci.com/github/qm012/go-http-routing-benchmark/builds/231071258 Thank you all for your reviews. |
|
@qm012 Please provide the benchmark between master and PR and show the result of |
|
@qm012 I make the PR https://github.com/qm012/go-http-routing-benchmark/pull/1/files to change the go version and update the testing count. Please update the new result by $ benchstat old.txt new.txt
name old time/op new time/op delta
Gin_Param 72.0ns ± 0% 73.4ns ± 0% ~ (p=1.000 n=1+1)
Gin_Param5 122ns ± 0% 146ns ± 0% ~ (p=1.000 n=1+1)
Gin_Param20 320ns ± 0% 432ns ± 0% ~ (p=1.000 n=1+1)
Gin_ParamWrite 127ns ± 0% 130ns ± 0% ~ (p=1.000 n=1+1)
Gin_GithubStatic 91.2ns ± 0% 130.0ns ± 0% ~ (p=1.000 n=1+1)
Gin_GithubParam 160ns ± 0% 238ns ± 0% ~ (p=1.000 n=1+1)
Gin_GithubAll 31.9µs ± 0% 48.1µs ± 0% ~ (p=1.000 n=1+1)
Gin_GPlusStatic 65.4ns ± 0% 95.3ns ± 0% ~ (p=1.000 n=1+1)
Gin_GPlusParam 93.7ns ± 0% 137.0ns ± 0% ~ (p=1.000 n=1+1)
Gin_GPlus2Params 119ns ± 0% 153ns ± 0% ~ (p=1.000 n=1+1)
Gin_GPlusAll 1.29µs ± 0% 1.73µs ± 0% ~ (p=1.000 n=1+1)
Gin_ParseStatic 66.0ns ± 0% 85.2ns ± 0% ~ (p=1.000 n=1+1)
Gin_ParseParam 75.8ns ± 0% 86.4ns ± 0% ~ (p=1.000 n=1+1)
Gin_Parse2Params 87.5ns ± 0% 103.0ns ± 0% ~ (p=1.000 n=1+1)
Gin_ParseAll 2.23µs ± 0% 2.92µs ± 0% ~ (p=1.000 n=1+1)
Gin_StaticAll 24.1µs ± 0% 36.2µs ± 0% ~ (p=1.000 n=1+1) |
|
hi, @appleboy Thanks you for reminding me. master https://travis-ci.com/github/qm012/go-http-routing-benchmark/builds/231194542 The $ benchstat old.txt new.txt
name old time/op new time/op delta
Gin_Param 91.1ns ± 2% 119.2ns ± 4% +30.92% (p=0.000 n=10+10)
Gin_Param5 163ns ± 2% 253ns ± 3% +54.70% (p=0.000 n=9+10)
Gin_Param20 399ns ± 2% 712ns ± 8% +78.57% (p=0.000 n=9+10)
Gin_ParamWrite 174ns ± 1% 219ns ± 5% +26.21% (p=0.000 n=10+9)
Gin_GithubStatic 129ns ± 1% 223ns ± 1% +72.08% (p=0.000 n=9+9)
Gin_GithubParam 229ns ± 2% 429ns ± 2% +87.68% (p=0.000 n=10+9)
Gin_GithubAll 42.8µs ± 2% 73.0µs ± 2% +70.56% (p=0.000 n=9+10)
Gin_GPlusStatic 91.5ns ± 2% 147.0ns ± 2% +60.60% (p=0.000 n=10+9)
Gin_GPlusParam 137ns ± 2% 209ns ± 1% +53.05% (p=0.000 n=10+8)
Gin_GPlus2Params 180ns ± 1% 286ns ± 2% +59.00% (p=0.000 n=9+9)
Gin_GPlusAll 1.82µs ± 2% 2.86µs ± 1% +57.30% (p=0.000 n=10+10)
Gin_ParseStatic 99.3ns ± 3% 146.9ns ± 4% +47.92% (p=0.000 n=10+10)
Gin_ParseParam 106ns ± 1% 143ns ± 1% +35.59% (p=0.000 n=9+8)
Gin_Parse2Params 127ns ± 2% 183ns ± 3% +43.87% (p=0.000 n=10+10)
Gin_ParseAll 3.15µs ± 2% 4.90µs ± 2% +55.56% (p=0.000 n=10+10)
Gin_StaticAll 31.6µs ± 1% 59.4µs ± 2% +87.94% (p=0.000 n=9+10) |
|
Benchstat report of the newly submitted Code: $ benchstat old.txt new.txt
name old time/op new time/op delta
Gin_Param 93.5ns ± 5% 99.5ns ± 0% +6.49% (p=0.000 n=10+7)
Gin_Param5 166ns ± 2% 216ns ± 1% +30.03% (p=0.000 n=10+9)
Gin_Param20 407ns ± 2% 612ns ± 1% +50.42% (p=0.000 n=10+9)
Gin_ParamWrite 177ns ± 2% 187ns ± 2% +5.51% (p=0.000 n=9+9)
Gin_GithubStatic 135ns ± 1% 193ns ± 2% +43.04% (p=0.000 n=9+10)
Gin_GithubParam 236ns ± 5% 367ns ± 1% +55.40% (p=0.000 n=10+9)
Gin_GithubAll 43.8µs ± 2% 62.1µs ± 3% +41.81% (p=0.000 n=10+9)
Gin_GPlusStatic 93.9ns ± 3% 123.6ns ± 2% +31.58% (p=0.000 n=9+9)
Gin_GPlusParam 139ns ± 1% 184ns ± 2% +31.82% (p=0.000 n=9+9)
Gin_GPlus2Params 183ns ± 1% 246ns ± 2% +34.41% (p=0.000 n=10+8)
Gin_GPlusAll 1.84µs ± 2% 2.49µs ± 2% +35.36% (p=0.000 n=10+9)
Gin_ParseStatic 93.2ns ± 1% 131.3ns ± 1% +40.90% (p=0.000 n=10+9)
Gin_ParseParam 108ns ± 2% 129ns ± 1% +20.13% (p=0.000 n=10+8)
Gin_Parse2Params 133ns ± 3% 160ns ± 1% +20.37% (p=0.000 n=9+9)
Gin_ParseAll 3.11µs ± 1% 4.32µs ± 2% +38.70% (p=0.000 n=8+10)
Gin_StaticAll 32.3µs ± 2% 51.2µs ± 1% +58.16% (p=0.000 n=10+10)I will continue to optimize the report, but the current result is not very ideal // match '/', If this condition is matched, the next route is found
if strings.HasSuffix(n.path, "/") || strings.Contains(n.fullPath, ":") || n.path == "" {
matchNum++
}Thank you all for your reviews.😆 |
|
The benchmarks: The name old time/op new time/op delta
Gin_Param 62.5ns ± 4% 64.0ns ± 5% ~ (p=0.110 n=10+10)
Gin_Param5 110ns ± 3% 114ns ± 5% +3.31% (p=0.017 n=10+10)
Gin_Param20 276ns ± 2% 298ns ± 3% +7.92% (p=0.000 n=10+10)
Gin_ParamWrite 118ns ± 2% 120ns ± 5% ~ (p=0.136 n=10+10)
Gin_GithubStatic 78.3ns ± 2% 96.4ns ± 2% +23.07% (p=0.000 n=10+10)
Gin_GithubParam 129ns ± 0% 177ns ± 2% +37.55% (p=0.000 n=10+10)
Gin_GithubAll 25.5µs ± 0% 32.1µs ± 1% +26.01% (p=0.000 n=10+10)
Gin_GPlusStatic 62.7ns ± 0% 69.5ns ± 2% +10.90% (p=0.000 n=10+10)
Gin_GPlusParam 88.6ns ± 3% 105.4ns ± 2% +18.98% (p=0.000 n=10+10)
Gin_GPlus2Params 112ns ± 2% 138ns ± 1% +23.81% (p=0.000 n=10+10)
Gin_GPlusAll 1.14µs ± 1% 1.38µs ± 1% +20.95% (p=0.000 n=10+9)
Gin_ParseStatic 59.2ns ± 0% 67.1ns ± 2% +13.25% (p=0.000 n=10+10)
Gin_ParseParam 70.6ns ± 4% 75.7ns ± 2% +7.22% (p=0.000 n=10+10)
Gin_Parse2Params 83.3ns ± 1% 90.8ns ± 1% +8.95% (p=0.000 n=9+10)
Gin_ParseAll 1.97µs ± 1% 2.27µs ± 1% +15.50% (p=0.000 n=10+10)
Gin_StaticAll 18.6µs ± 0% 25.0µs ± 1% +34.31% (p=0.000 n=10+10) |
add more comments update comment add example fix benchmark not found add comment and update test method gin_integration_test.go#L407 update comment and lastedNode directly assign current node optimize code Optimize the search next router logic optimize code Adjust the matching rules Adjust the matching order update condition code
|
hi @appleboy I updated the code, and when looking for the contents of the test report, I found that the relevant code submitted with (#2706) maintained two nodes, which would cause the test report to fail to get the expected results, so I made changes and optimizations. By maintaining a node, I controlled the way to find nodes, and made several cross-test reports for comparison when judging test conditions, and got an acceptable final result. Thank you for your help and review. The benchmarks:
The benchstat report: name old time/op new time/op delta
Gin_Param 63.5ns ± 4% 62.6ns ± 4% ~ (p=0.448 n=10+10)
Gin_Param5 112ns ± 6% 114ns ± 3% ~ (p=0.109 n=10+10)
Gin_Param20 282ns ± 3% 297ns ± 1% +5.47% (p=0.000 n=10+10)
Gin_ParamWrite 119ns ± 5% 118ns ± 2% ~ (p=0.566 n=10+10)
Gin_GithubStatic 81.2ns ± 6% 79.3ns ± 0% ~ (p=0.085 n=10+10)
Gin_GithubParam 128ns ± 6% 132ns ± 1% +2.64% (p=0.006 n=10+10)
Gin_GithubAll 26.2µs ± 3% 27.1µs ± 1% +3.10% (p=0.004 n=10+10)
Gin_GPlusStatic 66.0ns ± 3% 63.3ns ± 2% -4.17% (p=0.000 n=10+10)
Gin_GPlusParam 91.2ns ± 3% 86.1ns ± 1% -5.62% (p=0.000 n=10+10)
Gin_GPlus2Params 117ns ± 5% 110ns ± 1% -6.32% (p=0.000 n=9+10)
Gin_GPlusAll 1.23µs ± 7% 1.22µs ± 0% ~ (p=1.000 n=10+8)
Gin_ParseStatic 62.4ns ± 5% 61.9ns ± 1% ~ (p=0.218 n=10+10)
Gin_ParseParam 70.2ns ± 6% 68.3ns ± 1% ~ (p=0.184 n=10+10)
Gin_Parse2Params 82.5ns ± 4% 83.7ns ± 1% +1.49% (p=0.014 n=10+9)
Gin_ParseAll 2.01µs ± 3% 2.05µs ± 2% ~ (p=0.287 n=10+10)
Gin_StaticAll 18.9µs ± 3% 19.4µs ± 1% +2.93% (p=0.000 n=10+10) |
|
@g1eny0ung Please also take look at it. #2767 (comment) |
I'll take a look these few days. Sorry because I'm a little busy this week. 😵 |
|
hi @appleboy After a lot of cross-testing and comparison, the latest feedback test report is very exciting.I also tried to match '/' in another way, but the final result was not satisfactory. The latest benchmarks:
The latest benchstat report: name old time/op new time/op delta
Gin_Param 63.5ns ± 4% 60.6ns ± 1% -4.54% (p=0.000 n=10+10)
Gin_Param5 112ns ± 6% 103ns ± 1% -7.63% (p=0.000 n=10+9)
Gin_Param20 282ns ± 3% 259ns ± 3% -7.91% (p=0.000 n=10+10)
Gin_ParamWrite 119ns ± 5% 117ns ± 3% ~ (p=0.160 n=10+10)
Gin_GithubStatic 81.2ns ± 6% 71.9ns ± 2% -11.36% (p=0.000 n=10+10)
Gin_GithubParam 128ns ± 6% 119ns ± 1% -7.16% (p=0.000 n=10+10)
Gin_GithubAll 26.2µs ± 3% 25.3µs ± 1% -3.41% (p=0.000 n=10+10)
Gin_GPlusStatic 66.0ns ± 3% 58.2ns ± 2% -11.86% (p=0.000 n=10+10)
Gin_GPlusParam 91.2ns ± 3% 76.0ns ± 2% -16.70% (p=0.000 n=10+8)
Gin_GPlus2Params 117ns ± 5% 95ns ± 1% -19.17% (p=0.000 n=9+8)
Gin_GPlusAll 1.23µs ± 7% 1.09µs ± 2% -11.36% (p=0.000 n=10+10)
Gin_ParseStatic 62.4ns ± 5% 62.0ns ± 2% ~ (p=0.280 n=10+10)
Gin_ParseParam 70.2ns ± 6% 69.0ns ± 4% ~ (p=0.218 n=10+10)
Gin_Parse2Params 82.5ns ± 4% 83.0ns ± 2% ~ (p=0.271 n=10+10)
Gin_ParseAll 2.01µs ± 3% 1.94µs ± 2% -3.86% (p=0.008 n=10+10)
Gin_StaticAll 18.9µs ± 3% 18.4µs ± 1% -2.59% (p=0.005 n=10+10) |
Gin_Param5 112ns ± 6% 119ns ± 4% +6.74% (p=0.000 n=10+10)
Gin_Param20 282ns ± 3% 310ns ± 2% +10.09% (p=0.000 n=10+10)@rw-access @g1eny0ung any suggestion about this? |
|
hi @appleboy @rw-access @thinkerou I updated the code The latest benchmarks:
The latest benchstat report: name old time/op new time/op delta
Gin_Param 63.5ns ± 4% 61.3ns ± 1% -3.52% (p=0.001 n=10+10)
Gin_Param5 112ns ± 6% 116ns ± 1% +4.02% (p=0.002 n=10+9)
Gin_Param20 282ns ± 3% 305ns ± 1% +8.33% (p=0.000 n=10+9)
Gin_ParamWrite 119ns ± 5% 116ns ± 2% -2.97% (p=0.015 n=10+10)
Gin_GithubStatic 81.2ns ± 6% 80.3ns ± 0% ~ (p=0.172 n=10+10)
Gin_GithubParam 128ns ± 6% 136ns ± 1% +6.49% (p=0.000 n=10+10)
Gin_GithubAll 26.2µs ± 3% 28.1µs ± 0% +7.26% (p=0.000 n=10+10)
Gin_GPlusStatic 66.0ns ± 3% 61.4ns ± 0% -6.93% (p=0.000 n=10+10)
Gin_GPlusParam 91.2ns ± 3% 83.3ns ± 0% -8.65% (p=0.000 n=10+10)
Gin_GPlus2Params 117ns ± 5% 108ns ± 0% -8.00% (p=0.000 n=9+9)
Gin_GPlusAll 1.23µs ± 7% 1.19µs ± 0% ~ (p=0.464 n=10+10)
Gin_ParseStatic 62.4ns ± 5% 60.7ns ± 1% -2.75% (p=0.002 n=10+10)
Gin_ParseParam 70.2ns ± 6% 68.0ns ± 0% -3.08% (p=0.002 n=10+9)
Gin_Parse2Params 82.5ns ± 4% 85.8ns ± 0% +4.11% (p=0.000 n=10+9)
Gin_ParseAll 2.01µs ± 3% 1.99µs ± 0% ~ (p=0.466 n=10+10) |
|
hi erveryone,more balanced test reporting The latest benchmarks:
The latest benchstat report: name old time/op new time/op delta
Gin_Param 63.5ns ± 4% 61.4ns ± 5% -3.31% (p=0.019 n=10+10)
Gin_Param5 112ns ± 6% 115ns ± 3% +2.95% (p=0.012 n=10+10)
Gin_Param20 282ns ± 3% 304ns ± 2% +7.79% (p=0.000 n=10+10)
Gin_ParamWrite 119ns ± 5% 117ns ± 3% ~ (p=0.102 n=10+10)
Gin_GithubStatic 81.2ns ± 6% 78.4ns ± 2% -3.44% (p=0.009 n=10+10)
Gin_GithubParam 128ns ± 6% 133ns ± 1% +3.73% (p=0.001 n=10+10)
Gin_GithubAll 26.2µs ± 3% 27.1µs ± 0% +3.31% (p=0.004 n=10+10)
Gin_GPlusStatic 66.0ns ± 3% 62.7ns ± 2% -5.00% (p=0.000 n=10+10)
Gin_GPlusParam 91.2ns ± 3% 91.4ns ± 2% ~ (p=0.796 n=10+10)
Gin_GPlus2Params 117ns ± 5% 117ns ± 1% ~ (p=1.000 n=9+10)
Gin_GPlusAll 1.23µs ± 7% 1.19µs ± 1% ~ (p=0.565 n=10+10)
Gin_ParseStatic 62.4ns ± 5% 59.9ns ± 1% -3.92% (p=0.001 n=10+10)
Gin_ParseParam 70.2ns ± 6% 68.1ns ± 1% -2.88% (p=0.023 n=10+10)
Gin_Parse2Params 82.5ns ± 4% 83.8ns ± 1% ~ (p=0.089 n=10+10)
Gin_ParseAll 2.01µs ± 3% 1.96µs ± 1% -2.91% (p=0.003 n=10+10)
Gin_StaticAll 18.9µs ± 3% 19.1µs ± 0% ~ (p=0.156 n=10+9) |
|
LGTM, need @g1eny0ung @thinkerou @rw-access review and confirm. |
merge and bump v1.7.3? thanks! @appleboy |
|
@thinkerou we can't release the v1.7.3 version since of the issue #2786 @qm012 I can reproduce the issue with the latest commit d4ca9a0 package main
import (
"fmt"
"github.com/gin-gonic/gin"
)
func handler01(c *gin.Context) {
fmt.Println(c.Params)
c.String(200, "ok01")
}
func handler02(c *gin.Context) {
fmt.Println(c.Params)
c.String(200, "ok02")
}
func main() {
g := gin.Default()
g.GET("/get/abc", handler01)
g.GET("/get/:param", handler02)
g.Run(":9090")
}result: /abc 200 ok02
/a 404
/ab 200 ok02
/xyz 200 ok02
/abcd 200 ok02 |
@qm012 @appleboy Is it by design? |
|
@sdojjy No, the test cases have not been fully covered. Thank you for your feedback. |
|
this brought problems thus I am now frozen at #2878 |
(cherry picked from commit d4ca9a0)
(cherry picked from commit d4ca9a0)
content