Duplicate Json Parser

[codeZeilen]

With the dedicated Json package, we now have two Json parsers in the base system: Json and WebUtils.

I would suggest removing the WebUtils Json parsing logic and forwarding the requests to Json. There are minor differences between the two parsers. One of the most noticable I discovered so far is how invalid input is handled (Json returns nil, WebUtils throws an Error).

[krono]

Am 22. August 2023 08:16:16 MESZ schrieb Patrick R @.***>:

With the dedicated Json package, we now have two Json parsers in the base system: Json and WebUtils.

I would suggest removing the WebUtils Json parsing logic and forwarding the requests to Json. There are minor differences between the two parsers. One of the most noticable I discovered so far is how invalid input is handled (Json returns nil, WebUtils throws an Error).

do we have perf numbers for both? -- Sent from a mobile device

[codeZeilen]

First simple benchmark suggests Json is significantly faster.

input := '{
    "benchmark_name": "Beispiel-Benchmark",
    "system_info": {
        "os": "Windows 10",
        "cpu": "Intel Core i7-8700K",
        "ram": "16 GB"
    },
    "test_results": [
        {
            "test_name": "Test 1",
            "duration_ms": 1234,
            "result": "pass"
        },
        {
            "test_name": "Test 2",
            "duration_ms": 5678,
            "result": "fail"
        }
    ]
}
'.
object := WebUtils jsonDecode: input readStream.

"DECODE"
[Json readFrom: input readStream] bench  
'128,000 per second. 7.81 microseconds per run. 2.23955 % GC time.' 
'128,000 per second. 7.78 microseconds per run. 2.05959 % GC time.' 

[WebUtils jsonDecode: input readStream] bench  
'96,800 per second. 10.3 microseconds per run. 2.61948 % GC time.' 
'92,700 per second. 10.8 microseconds per run. 2.71946 % GC time.' 

"ENCODE"
[Json render: object] bench 
'357,000 per second. 2.8 microseconds per run. 7.71846 % GC time.' 
'360,000 per second. 2.78 microseconds per run. 7.69692 % GC time.' 

[WebUtils jsonEncode: object] bench  
'115,000 per second. 8.73 microseconds per run. 10.77784 % GC time.' 
'114,000 per second. 8.8 microseconds per run. 10.69572 % GC time.' ```

[krono]

Am 22. August 2023 14:56:56 MESZ schrieb Patrick R @.***>:

First simple benchmark suggests Json is significantly faster.

input := '{
   "benchmark_name": "Beispiel-Benchmark",
   "system_info": {
       "os": "Windows 10",
       "cpu": "Intel Core i7-8700K",
       "ram": "16 GB"
   },
   "test_results": [
       {
           "test_name": "Test 1",
           "duration_ms": 1234,
           "result": "pass"
       },
       {
           "test_name": "Test 2",
           "duration_ms": 5678,
           "result": "fail"
       }
   ]
}
'.
[Json readFrom: input readStream] bench  

'128,000 per second. 7.81 microseconds per run. 2.23955 % GC time.' 
'128,000 per second. 7.78 microseconds per run. 2.05959 % GC time.' 
[WebUtils jsonDecode: input readStream] bench  

'96,800 per second. 10.3 microseconds per run. 2.61948 % GC time.' 
'92,700 per second. 10.8 microseconds per run. 2.71946 % GC time.' 

object := WebUtils jsonDecode: input readStream.

[Json render: object] bench 
'357,000 per second. 2.8 microseconds per run. 7.71846 % GC time.' 
'360,000 per second. 2.78 microseconds per run. 7.69692 % GC time.' 

[WebUtils jsonEncode: object] bench  
'115,000 per second. 8.73 microseconds per run. 10.77784 % GC time.' 
'114,000 per second. 8.8 microseconds per run. 10.69572 % GC time.' ```

neat!

at its inception, WebClient was meant to also run on other Smalltalks (Pharo, for example), hence webutils.

but since this is now trunk, nuke WebClient Json. -T -- Sent from a mobile device

[marceltaeumel]

I would suggest removing the WebUtils Json parsing logic and forwarding the requests to Json.

+1