Skip to main content
Huy's site

jq is a powerful JSON processor

So lately I’ve been using jq quite a bit. It is a CLI JSON processor that makes your life easier if you have to deal with a lot of JSON. Here I’m going to give two examples of how it’s used.

Search response

I’ve been dealing with JSON from Elasticsearch API1, but as they would not release their documents under a free license, I will take an example from OpenSearch instead2:

{
  "took": 39,
  "timed_out": false,
  "_shards": {
    "total": 68,
    "successful": 68,
    "skipped": 0,
    "failed": 0
  },
  "hits": {
    "total": {
      "value": 5837,
      "relation": "eq"
    },
    "max_score": 7.8623476,
    "hits": [
      {
        "_index": "new_shakespeare",
        "_type": "_doc",
        "_id": "100763",
        "_score": 7.8623476,
        "_source": {
          "type": "line",
          "line_id": 100764,
          "play_name": "Troilus and Cressida",
          "speech_number": 43,
          "line_number": "3.1.68",
          "speaker": "PANDARUS",
          "text_entry": "Sweet queen, sweet queen! thats a sweet queen, i faith."
        }
      },
      {
        "_index": "shakespeare",
        "_type": "_doc",
        "_id": "28559",
        "_score": 5.8923807,
        "_source": {
          "type": "line",
          "line_id": 28560,
          "play_name": "Cymbeline",
          "speech_number": 20,
          "line_number": "1.1.81",
          "speaker": "QUEEN",
          "text_entry": "No, be assured you shall not find me, daughter,"
        }
      }
    ]
  }
}

Woah that’s long!

How do you know how much time it took? I want to measure the performance.

$ curl ... | jq '.took'
39

Nice. Not very helpful, though—I can easily see that at the top of the result. I want to see which responses were returned.

$ curl ... | jq '.hits.hits'
[
  {
    "_index": "new_shakespeare",
    "_type": "_doc",
    "_id": "100763",
    "_score": 7.8623476,
    "_source": {
      "type": "line",
      "line_id": 100764,
      "play_name": "Troilus and Cressida",
      "speech_number": 43,
      "line_number": "3.1.68",
      "speaker": "PANDARUS",
      "text_entry": "Sweet queen, sweet queen! thats a sweet queen, i faith."
    }
  },
  {
    "_index": "shakespeare",
    "_type": "_doc",
    "_id": "28559",
    "_score": 5.8923807,
    "_source": {
      "type": "line",
      "line_id": 28560,
      "play_name": "Cymbeline",
      "speech_number": 20,
      "line_number": "1.1.81",
      "speaker": "QUEEN",
      "text_entry": "No, be assured you shall not find me, daughter,"
    }
  }
]

The response is the value of hits inside a hits, so the query is .hits.hits. How intuitive!

But this is hard to read. Let’s just take the play_name of the result.

$ curl ... | jq .hits.hits[]._source.play_name
"Troilus and Cressida"
"Cymbeline"

Neat! But how does it work? The brackets [] signifies that we should take all the results, from each of which the value for ._source.play_name is taken.

There might be the option to write that in the search query DSL, which also reduces the data transferred, but I bet this is much easier to write.

Wallpaper

I use feh for setting wallpaper, which can take an image from the internet. Previously, I used a static list that I collected myself, but I recently discovered a wallpaper API from wallhaven.

Let’s say, I want to get a space image as wallpaper, I would use this command:

curl -s 'https://wallhaven.cc/api/v1/search?q=space&ratios=16x9&sorting=toplist'

Where q is the query, ratios is the image ratio so that it fits the screen, and sorting is the way the results would be sorted before pagination. The results is long, but it looks like this:

{
  "data": [
      {
      "id": "l3zmwy",
      "url": "https://wallhaven.cc/w/l3zmwy",
      "short_url": "https://whvn.cc/l3zmwy",
      "views": 67050,
      "favorites": 705,
      "source": "https://www.artstation.com/artwork/YaQwgP",
      "purity": "sfw",
      "category": "general",
      "dimension_x": 1920,
      "dimension_y": 1080,
      "resolution": "1920x1080",
      "ratio": "1.78",
      "file_size": 781731,
      "file_type": "image/jpeg",
      "created_at": "2021-05-18 19:26:23",
      "colors": [
        "#424153",
        "#000000",
        "#663300",
        "#996633",
        "#999999"
      ],
      "path": "https://w.wallhaven.cc/full/l3/wallhaven-l3zmwy.jpg",
      "thumbs": {
        "large": "https://th.wallhaven.cc/lg/l3/l3zmwy.jpg",
        "original": "https://th.wallhaven.cc/orig/l3/l3zmwy.jpg",
        "small": "https://th.wallhaven.cc/small/l3/l3zmwy.jpg"
      }
    },
    ...
  ]
}

So, to get the image path from this, I run:

curl ... | jq -r .data[].path | shuf -n 1 | feh --bg-center

I use shuf because I’d like a new wallpaper every time I run this script. Put this as the startup script or add a cron job and you’ll have a changing wallpaper. Disclaimer: this script does not always work, because of feh. If you’re using GNOME, for example, feh can’t be used to set background.


  1. Yes I know it’s no longer free software (it still partially is I think). I have no choice, and I would still be more at peace with a source available software that is self-hostable. ↩︎

  2. Copyright 2021 OpenSearch contributors. Released under Apache License. ↩︎


Would you like to discuss this post? Email me!

Fediring

Look at my fedi fellows' sites:
  1. Previous site
  2. What is Fediring?
  3. Next site

Articles from blogs I read

A Hare code generator for finding ioctl numbers

Modern Unix derivatives have this really bad idea called ioctl. It’s a function which performs arbit…

via Drew DeVault's blog May 14, 2022

Phone-tracking companies tracking military and intelligence personnel

Two US phone-tracking companies showed they could track military and intelligence personnel of …

via Richard Stallman's Political Notes May 13, 2022

GCC 12 Becoming Default Compiler in Tumbleweed

More than a month after preparing the default compiler for openSUSE Tumbleweed to be switched to GN…

via openSUSE News May 13, 2022
Generated by openring