[SOLVED] How to put specific filenames into a specific JSON format using bash or Perl?

Issue

Assuming I'm in the folder like this:

➜  tmp.lDrLPUOF ls
1.txt 2.txt 3.txt 1.zip 2.rb

I want to put all the filenames of text files into a specific JSON format like this:

{
  "": [
    {
      "title": "",
      "file": "1"
    },
    {
      "title": "",
      "file": "2"
    },
    {
      "title": "",
      "file": "3"
    }
  ]
}

Now I just know how to list all the filenames:

➜  tmp.lDrLPUOF ls *'.txt'
1.txt 2.txt 3.txt

Can I use bash or Perl to achieve this purpose? Thank you very much!

Edit

Thanks for @Charles Duffy and @Shawn 's great answers. But it's my fault to forget another important piece of information——time. I want to put the filenames into the resulting JSON per their creating time.

The creating time is as below:

➜  tmp.lDrLPUOF ls -lTr
total 0
-rw-r--r--  1 administrator  staff  0 Oct 12 09:35:05 2022 3.txt
-rw-r--r--  1 administrator  staff  0 Oct 12 09:35:08 2022 2.txt
-rw-r--r--  1 administrator  staff  0 Oct 12 09:35:12 2022 1.txt

So the resulting JSON I wanted should be like this:

{
  "": [
    {
      "title": "",
      "file": "3"
    },
    {
      "title": "",
      "file": "2"
    },
    {
      "title": "",
      "file": "1"
    }
  ]
}

Solution

{ shopt -s nullglob; set -- *.txt; printf '%s\0' "$@"; } | jq -Rn '
  {"": [ input
         | split("\u0000")[]
         | select(. != "")
         | {"title": "",
            "file": . | rtrimstr(".txt")
           }
       ]
  }
'

Let's break this down into pieces.

On the bash side:

• shopt -s nullglob tells the shell that if *.txt has no arguments, it should emit nothing at all, instead of emitting the string *.txt as a result.
• set -- overwrites the argument list in the current context (because this is a block on the left-hand side of the pipeline that context is transient and won't change "$@" in code outside the pipe).
• printf '%s\0' "$@" prints our arguments, with a NUL character after each one; if there are no arguments at all, it prints only a NUL.

On the jq side:

• -R specifies that the input is raw data, not json.
• -n specifies that we don't automatically consume any inputs, but will instead use input or inputs to specify where input should be read.
• split("\u0000") splits the input on NULs. (This is important because the NUL is the only character that can't exist in a filename, which is why we used printf '%s\0' on the shell end; that way we work correctly with filenames with newlines, literal quotes, whitespace, and all the other weirdness that's able to exist).
• select(. != "") ignores empty strings.
• rtrimstr(".txt") removes .txt from the name.

---

Addendum: Sorting by mtime

The jq parts don't need to be modified here: to sort by mtime you can adjust only the shell. On a system with GNU find, sort and sed, this might look like:

find . -maxdepth 1 -type f -name '*.txt' -printf '%T@ %P\0' |
  sort -zn |
  sed -z -re 's/^[[:digit:].]+ //g' |
  jq -Rn '

...followed by the same jq given above.

Answered By - Charles Duffy

Answer Checked By - Marilyn (WPSolving Volunteer)