github.com/unigraph-dev/dgraph@v1.1.1-0.20200923154953-8b52b426f765/gql/README.md (about) 1 Comparing the old and new methods, we find that using slice makes the parsing 20% faster on 2 average than using channels. Also, using slices allows the parser to backtrack and peek the 3 tokens which couldn't be done using channels as each token can only be consumed once. 4 5 ``` 6 Name unit Old New Improvement 7 ---------------------------------------------------------------------- 8 Benchmark_Filters-4 ns/op 14007 9634 31 % 9 Benchmark_Geq-4 ns/op 11701 8602 26 % 10 Benchmark_Date-4 ns/op 11687 8630 26 % 11 Benchmark_directors-4 ns/op 18663 14201 23 % 12 Benchmark_Filters_parallel-4 ns/op 6486 5015 22 % 13 Benchmark_Movies-4 ns/op 16097 12807 20 % 14 Benchmark_directors_parallel-4 ns/op 8766 6966 20 % 15 Benchmark_Mutation-4 ns/op 5167 4155 19 % 16 Benchmark_Movies_parallel-4 ns/op 7537 6151 18 % 17 Benchmark_Date_parallel-4 ns/op 5462 4515 17 % 18 Benchmark_Geq_parallel-4 ns/op 5390 4485 16 % 19 Benchmark_Mutation_parallel-4 ns/op 2326 2161 07 % 20 Benchmark_Mutation1000-4 ns/op 549428 512851 06 % 21 Benchmark_Mutation1000_parallel-4 ns/op 261785 254911 02 % 22 ```