Feature request

[Albonycal]

Yesterday I was testing intel_gpu_top utility in linux which gives some information & has an option to output json with -J flag I wanted to write a script using this functionality using gron.. That utility constantly outputs in json which looks like this:

    "period": {
        "duration": 1000.323144,
        "unit": "ms"
    },
    "frequency": {
        "requested": 649.790024,
        "actual": 649.790024,
        "unit": "MHz"
    },
    "interrupts": {
        "count": 8.997093,
        "unit": "irq/s"
    },
    "rc6": {
        "value": 85.332937,
        "unit": "%"
    },
    "power": {
        "value": 0.192778,
        "unit": "W"
    },
    "engines": {
        "Render/3D/0": {
            "busy": 0.000000,
            "sema": 0.000000,
            "wait": 0.000000,
            "unit": "%"
        },
        "Blitter/0": {
            "busy": 0.000000,
            "sema": 0.000000,
            "wait": 0.000000,
            "unit": "%"
        },
        "Video/0": {
            "busy": 0.000000,
            "sema": 0.000000,
            "wait": 0.000000,
            "unit": "%"
        }
    }
}

My problem with is that it stops after reading the first value outputted by the tool. then the process stops... I want it to constantly show the output.. Example:

json = {};
json.engines = {};
json.engines["Blitter/0"] = {};
json.engines["Blitter/0"].busy = 0.000000;
json.engines["Blitter/0"].sema = 0.000000;
json.engines["Blitter/0"].unit = "%";
json.engines["Blitter/0"].wait = 0.000000;
json.engines["Render/3D/0"] = {};
json.engines["Render/3D/0"].busy = 0.000000;
json.engines["Render/3D/0"].sema = 0.000000;
json.engines["Render/3D/0"].unit = "%";
json.engines["Render/3D/0"].wait = 0.000000;
json.engines["Video/0"] = {};
json.engines["Video/0"].busy = 0.000000;
json.engines["Video/0"].sema = 0.000000;
json.engines["Video/0"].unit = "%";
json.engines["Video/0"].wait = 0.000000;
json.frequency = {};
json.frequency.actual = 0.000000;
json.frequency.requested = 0.000000;
json.frequency.unit = "MHz";
json.interrupts = {};
json.interrupts.count = 0.000000;
json.interrupts.unit = "irq/s";
json.period = {};
json.period.duration = 0.007298;
json.period.unit = "ms";
json.power = {};
json.power.unit = "W";
json.power.value = 0.000000;
json.rc6 = {};
json.rc6.unit = "%";
json.rc6.value = 0.000000;

It formats the output only one time and the stops but I want to do it constantly Is their some shell trick that can be used to do this (or a new feature) Thank you @albonycal

[cho-m]

Disclaimer: I'm not a shell expert, but I'll give some options I can think of.

Based on gron's current features, one possible option would be to process each separate JSON object into a single line and then use gron's streaming feature.

I can't comment on if this will work for intel_gpu_top, but the consolidation could be done with something like jq -c option for something like:

<process generating output> | jq --unbuffered -c '.' | gron -s

This will represent each JSON object as a value inside a JSON array, so a continuously running process can lead to very large number array index.

If you don't want the wrapping array, another possible option would be to send jq output to a loop that would repeatedly run gron. There may be some issues in my exact example, but hopefully the general gist comes across

<process generating output> | jq --unbuffered -c '.' | while read -r line; do echo "$line" | gron; done

---

Not too sure if there is a shell trick to restart gron after it exits. You will need to wait for someone else to chime in on that.

---

For gron feature, that doesn't rely on jq, it sounds like the option 2 that was not implemented from https://github.com/tomnomnom/gron/issues/23#issuecomment-259271490:

I've been giving some thought about the approach needed for this, and there's probably only two sane options:

1. Require that the input be one JSON blob per line so it's easy to split on \n
2. Do a pre-parse step to detect multiple JSON objects in the input

Option 1 is the implemented gron -s feature previously discussed.

[Albonycal]

nah it doesn't work.. anyway.. I stopped working on the script

[dotcs]

Not sure if I understood the question right, but maybe watch might help here?

E.g. place this script in /tmp/script.sh and give it executable rights. The script reads the uptime of the computer and creates a simple JSON document with the help of jq. Of course any additional steps, such as filtering the output and re-creating a JSON document with ungron could be done here as well.

#!/bin/bash
jq -c --null-input --arg uptime $(uptime | awk -F' ' '{ print $1 }') '{ uptime: $uptime }' | gron

Then run watch -n 1 /tmp/script.sh. The output is updated every 1 second.