feat(pkg/parser): impl expression eval

[WenyXu]

No description provided.

[WenyXu]

I compared the performance to the govaluate.

For the parsing performance, our implementation is about 20x-25x slower than the govaluate. In most cases, it's acceptable. We only parse the expression once when the expression is defined, and then the expression will be stored as an abstract syntax tree in the memory.

For the evaluation performance, In our scenarios, our implementation is about 2-8x faster than the govaluate. Notably, in the evaluation with parameters and functions, our implementation's performance is about 8x faster than the govaluate.

goos: linux
goarch: amd64
pkg: github.com/casbin-mesh/neo/pkg/parser
cpu: 12th Gen Intel(R) Core(TM) i7-12700
BenchmarkSingleParse_govaluate
BenchmarkSingleParse_govaluate-20                      	 1859112	       614.8 ns/op
BenchmarkSingleParse
BenchmarkSingleParse-20                                	  224336	     12154 ns/op
BenchmarkEvaluationSingle_govaluate
BenchmarkEvaluationSingle_govaluate-20                 	126624380	         8.701 ns/op
BenchmarkEvaluationSingle
BenchmarkEvaluationSingle-20                           	890921866	         1.361 ns/op
BenchmarkEvaluationNumericLiteral_govaluate
BenchmarkEvaluationNumericLiteral_govaluate-20         	30109729	        33.80 ns/op
BenchmarkEvaluationNumericLiteral
BenchmarkEvaluationNumericLiteral-20                   	121256878	         9.713 ns/op
BenchmarkEvaluationLiteralModifiers_govaluate
BenchmarkEvaluationLiteralModifiers_govaluate-20       	14240113	        83.52 ns/op
BenchmarkEvaluationLiteralModifiers
BenchmarkEvaluationLiteralModifiers-20                 	68422755	        15.39 ns/op
BenchmarkEvaluationParameters_govaluate
BenchmarkEvaluationParameters_govaluate-20             	17665630	        67.55 ns/op
BenchmarkEvaluationParameters
BenchmarkEvaluationParameters-20                       	37160481	        30.01 ns/op
BenchmarkEvaluationParametersModifiers_govaluate
BenchmarkEvaluationParametersModifiers_govaluate-20    	 9849256	       135.3 ns/op
BenchmarkEvaluationParametersModifiers
BenchmarkEvaluationParametersModifiers-20              	23305705	        51.33 ns/op
BenchmarkEvaluationFunction_govaluate
BenchmarkEvaluationFunction_govaluate-20               	 2959526	       405.5 ns/op
BenchmarkEvaluationFunction
BenchmarkEvaluationFunction-20                         	23955829	        53.09 ns/op

[WenyXu]

Update Benchmarks

I refactored the expression evaluation and fixed a bug that the code might mutate the value in the context during the expression's evaluation. I fixed that by introducing extra memory allocation during the evaluation, but it also reduced the performance.

BenchmarkEvaluationParametersModifiers_govaluate-20    	 7920292	       134.4 ns/op
BenchmarkEvaluationParametersModifiers-20              	 5718110	       185.6 ns/op

As you can see, when the evaluation only involves parameters, the performance was slightly worse than the govalute because it spent about 20% CPU time to allocate memory and copy the primitive's value.

BenchmarkEvaluationFunction_govaluate-20               	 2948817	       410.3 ns/op
BenchmarkEvaluationFunction-20                         	13911880	        92.45 ns/op
BenchmarkEvaluationFunctionOnly-20                     	19744444	        60.46 ns/op
BenchmarkEvaluationFunction_Naive-20                   	1000000000	         0.1200 ns/op

However, we still can be benefited in some scenarios. If the evaluation only involves functions, there are no allocation or copy operations.

goos: linux
goarch: amd64
pkg: github.com/casbin-mesh/neo/pkg/expression
cpu: 12th Gen Intel(R) Core(TM) i7-12700
BenchmarkSingleParse_govaluate
BenchmarkSingleParse_govaluate-20                      	 2228401	       619.7 ns/op
BenchmarkSingleParse
BenchmarkSingleParse-20                                	  126939	     13510 ns/op
BenchmarkEvaluationSingle_govaluate
BenchmarkEvaluationSingle_govaluate-20                 	131851587	         9.153 ns/op
BenchmarkEvaluationSingle
BenchmarkEvaluationSingle-20                           	770851652	         1.405 ns/op
BenchmarkEvaluationNumericLiteral_govaluate
BenchmarkEvaluationNumericLiteral_govaluate-20         	34579762	        35.61 ns/op
BenchmarkEvaluationNumericLiteral
BenchmarkEvaluationNumericLiteral-20                   	124814935	         9.716 ns/op
BenchmarkEvaluationLiteralModifiers_govaluate
BenchmarkEvaluationLiteralModifiers_govaluate-20       	14283465	        82.40 ns/op
BenchmarkEvaluationLiteralModifiers
BenchmarkEvaluationLiteralModifiers-20                 	74358448	        15.80 ns/op
BenchmarkEvaluationParameters_govaluate
BenchmarkEvaluationParameters_govaluate-20             	19192678	        62.81 ns/op
BenchmarkEvaluationParameters
BenchmarkEvaluationParameters-20                       	 9405704	       128.8 ns/op
BenchmarkEvaluationParametersModifiers_govaluate
BenchmarkEvaluationParametersModifiers_govaluate-20    	 7920292	       134.4 ns/op
BenchmarkEvaluationParametersModifiers
BenchmarkEvaluationParametersModifiers-20              	 5718110	       185.6 ns/op
BenchmarkEvaluationFunction_govaluate
BenchmarkEvaluationFunction_govaluate-20               	 2948817	       410.3 ns/op
BenchmarkEvaluationFunction
BenchmarkEvaluationFunction-20                         	13911880	        92.45 ns/op
BenchmarkEvaluationFunction_Naive
BenchmarkEvaluationFunction_Naive-20                   	1000000000	         0.1200 ns/op
BenchmarkEvaluationFunctionOnly
BenchmarkEvaluationFunctionOnly-20                     	19744444	        60.46 ns/op

[hsluoyz]

@nodece @noneback plz review