jsonsam 0.1.1

Tool to facilitate editing, splitting and merging of nested JSON and Python dictionaries

JSONSAM Overview

Quickstart

The following is a minimal example of the JSONSAM CLI workflow:

$ jsonsam my.json
Make edits to my-denorm.json and then rerun with modified file to generate updated JSON output
$ vim my-denorm.json 		# Add/delete/edit line-based paths
$ jsonsam my-denorm.json 	# Convert edits back to standard JSON
Updated JSON file written to my-denorm-norm.json

The finished product (the standard JSON file with edits applied) is my-denorm-norm.json. See the Command Line Examples section for extensive additional examples.

Introduction

JSONSAM captures some of the techniques I've used over the years to slice and dice large chunks (> 1MB) of highly nested JSON. vim is still my preferred editor and I like to use line-based shell primitives like sed and grep when possible. However, since conventional JSON is not line-framed and JSON-streaming has multipath semantics, it is generally not amenable to direct modification using conventional Unix utilities.

JSONSAM addresses this by converting JSON into an intermediate denormalized format (not to be confused with JSON-streaming or new-line delimited JSON). The denormalized format captures each path in a JSON list of line-delimited paths. These paths can then be edited with standard tools to add, delete, modify, shift-up/down, reorder nested lists, etc. Once desired edits are made, the denormalized file is renormalized back into a standard JSON file. JSONSAM can also work on entire directories of files at once, allowing them to be easily merged, split and modified. Denormalized form might also be useful for streaming if one is willing to trade data expansion for framing.

JSONSAM can also perform set arithmetic (union, intersection and difference) against a pair of JSON input files. This is useful when a template or reference file exists and it's desired to extract, remove or augment another JSON file based on this template.

Lastly, a random realistic dictionary generator is included as part of the test infrastructure. It creates nested dictionaries and lists of random breadth and width.

JSONSAM comprises the following components:

• jsonsam: Command line utility for processing JSON
• randdict: Command line utility for generating random dictionaries
• JsonSam: Denormalize, edit, normalize and perform set arithmetic on JSON files
• DictSam: Denormalize, edit, normalize and perform set arithmetic on nested dictionaries/lists
• DictGen: Generate random deeply nested dictionaries and lists with realistic data

Installation

Install from PyPi:

pip install jsonsam

Importing

>>> from jsonsam import DictSam # Dictionary denorm, norm, and set arithmetic
>>> from jsonsam import DictGen # Random realistic nested dictionary/list generator

Command Line Examples

# online help
$ jsonsam -h

Consider the following file simple.json:

$ cat simple.json
{
  "Bad": true,
  "beer": 3.14,
  "rots": "Our",
  "young": {
    "guts": "but",
    "Vodka": [
      27,
      [
        "goes",
        2.718,
        {
          "Well": "water",
          "max": "plank"
        },
        "old"
      ]
    ],
    "bark": null,
    "last": 66
  }
}

This file is in normalized form. First, we denormalize the file:

$ jsonsam simple.json
Make edits to simple-denorm.json and then rerun with modified file to generate updated JSON output

The denormalized file, simple-denorm.json, now contains a list of paths. Each line represents a path from root to leaf through the original JSON file with non-terminal integers representing list indices:

$ cat simple-denorm.json
["Bad", true]
["beer", 3.14]
["rots", "Our"]
["young", "Vodka", 0, 27]
["young", "Vodka", 1, 0, "goes"]
["young", "Vodka", 1, 1, 2.718]
["young", "Vodka", 1, 2, "Well", "water"]
["young", "Vodka", 1, 2, "max", "plank"]
["young", "Vodka", 1, 3, "old"]
["young", "bark", null]
["young", "guts", "but"]
["young", "last", 66]

All other rows can be arbitrarily mutated, added, removed and reordered (non-terminal integers denote nested list indices). When the desired result is achieved, rerun jsonsam with the edited simple-denorm.json as input to generate a renormalized simple-denorm-norm.json output file. This represents the original JSON file with the desired additions, deletions and mutations applied. Note that JSONSAM appends -norm and -denorm strings to the file stem for convenience to avoid overwriting original files and track lineage. This behavior can be overridden with the -o <outputfile> option.

Unless otherwise noted all examples that follow will start from simple.json and simple-denorm.json.

Relocate dictionary slice from nested lists

Suppose we want to shift-up the pair "Well": "water" out of the dictionary nested in the list of lists. This can be achieved by applying the following single-line diff to simple-denorm.json:

 ["young", "Vodka", 0, 27]
 ["young", "Vodka", 1, 0, "goes"]
 ["young", "Vodka", 1, 1, 2.718]
-["young", "Vodka", 1, 2, "Well", "water"]
+["young", "Vodka", 2, "Well", "water"]
 ["young", "Vodka", 1, 2, "max", "plank"]
 ["young", "Vodka", 1, 3, "old"]
 ["young", "bark", null]

To apply the changes, we normalize the path list by rerunning JSONSAM:

$ jsonsam simple-denorm.json 
Updated JSON file written to simple-denorm-norm.json
$ diff --unified=20 <(cat simple.json |jq --sort-keys '.') <(cat simple-denorm-norm.json |jq --sort-keys '.')
 {
   "Bad": true,
   "beer": 3.14,
   "rots": "Our",
   "young": {
     "Vodka": [
       27,
       [
         "goes",
         2.718,
         {
-          "Well": "water",
           "max": "plank"
         },
         "old"
-      ]
+      ],
+      {
+        "Well": "water"
+      }
     ],
     "bark": null,
     "guts": "but",
     "last": 66
   }
 }

We see in the above JSON that indeed the "Well": "water" pair has been plucked from the nested dictionary, and a new dictionary containing the pair has been appended to the outer list. simple-denorm-norm.json now represents the finished product.

Click to expand additional examples:

Remove an item, and convert a leaf string to a dictionary

Remove an item, and convert a leaf string to a dictionary

In this case we do two things:

1. Remove the young.bark.null path
2. Convert young.guts.but path from a terminal string to a dictionary of young.guts.{but:dog, route:cat}

This can be achieved by applying the following diff to simple-denorm.json:

 ["young", "Vodka", 1, 2, "Well", "water"]
 ["young", "Vodka", 1, 2, "max", "plank"]
 ["young", "Vodka", 1, 3, "old"]
-["young", "bark", null]
-["young", "guts", "but"]
+["young", "guts", "but", "dog"]
+["young", "guts", "route", "cat"]
 ["young", "last", 66]

Looking at simple-denorm-norm.json we see young.bark.null has been removed and we've converted young.guts.but from a terminal string into a dictionary of {but:dog, route:cat}:

$ jsonsam simple-denorm.json
Updated JSON file written to simple-denorm-norm.json
$ cat simple-denorm-norm.json
{
  "Bad": true,
  "beer": 3.14,
  "rots": "Our",
  "young": {
    "Vodka": [
      27,
      [
        "goes",
        2.718,
        {
          "Well": "water",
          "max": "plank"
        },
        "old"
      ]
    ],
    "guts": {
      "but": "dog",
      "route": "cat"
    },
    "last": 66
  }
}

Convert a terminal string to a list of dictionaries

Convert a terminal string to a list of dictionaries

Here we convert the young.guts path to a list of dictionaries. This can be achieved by applying the following diff to simple-denorm.json:

 ["young", "Vodka", 1, 2, "max", "plank"]
 ["young", "Vodka", 1, 3, "old"]
 ["young", "bark", null]
-["young", "guts", "but"]
+["young", "guts", 0, "but", "dog"]
+["young", "guts", 1, "route", "dog"]
 ["young", "last", 66]

And the resulting normalized JSON:

$ jsonsam simple-denorm.json
Updated JSON file written to simple-denorm-norm.json
$ cat simple-denorm-norm.json
{
  "Bad": true,
  "beer": 3.14,
  "rots": "Our",
  "young": {
    "Vodka": [
      27,
      [
        "goes",
        2.718,
        {
          "Well": "water",
          "max": "plank"
        },
        "old"
      ]
    ],
    "bark": null,
    "guts": [
      {
        "but": "dog"
      },
      {
        "route": "dog"
      }
    ],
    "last": 66
  }
}

Add path and reorder a list

Add path and reorder a list

Three tasks here:

1. Add a new jack.beanstock.happy.941 path
2. Move list item young.Vodka.1.2 to young.Vodka.2
3. Add the string "append" to young.Vodka list

This can be achieved by applying the following diff to simple-denorm.json:

 ["Bad", true]
 ["beer", 3.14]
 ["rots", "Our"]
+["jack", "beanstock", "happy", 941]
 ["young", "Vodka", 0, 27]
-["young", "Vodka", 1, 0, "goes"]
-["young", "Vodka", 1, 1, 2.718]
-["young", "Vodka", 1, 2, "Well", "water"]
-["young", "Vodka", 1, 2, "max", "plank"]
+["young", "Vodka", 1, 1, "goes"]
+["young", "Vodka", 1, 0, 2.718]
+["young", "Vodka", 2, "Well", "water"]
+["young", "Vodka", 2, "max", "plank"]
+["young", "Vodka", 3, "appended"]
 ["young", "Vodka", 1, 3, "old"]
 ["young", "bark", null]
 ["young", "guts", "but"]

Which results in:

$ jsonsam simple-denorm.json
Updated JSON file written to simple-denorm-norm.json
$ cat simple-denorm-norm.json
{
  "Bad": true,
  "beer": 3.14,
  "jack": {
    "beanstock": {
      "happy": 941
    }
  },
  "rots": "Our",
  "young": {
    "Vodka": [
      27,
      [
        2.718,
        "goes",
        null,
        "old"
      ],
      {
        "Well": "water",
        "max": "plank"
      },
      "appended"
    ],
    "bark": null,
    "guts": "but",
    "last": 66
  }
}

Removed list elements become null to preserve the ordinal of retained elements.

Unix pipeline processing

Unix pipeline processing

If no input files are provided, jsonsam defaults to reading from stdin and writing to stdout. This allows use of standard Unix pipelines for editing JSON files with line-based shell utilities like sed, grep, awk, etc.

In the below example, we use grep to remove all Vodka keys and sed to convert beer to Whiskey:

$ cat simple.json |jsonsam |grep -v Vodka |sed 's/beer/Whiskey/g' |jsonsam
{
  "Bad": true,
  "Whiskey": 3.14,
  "rots": "Our",
  "young": {
    "bark": null,
    "guts": "but",
    "last": 66
  }
}

Transmogrify 7 JSON files into 11 JSON files, round-robin paths

Transmogrify 7 JSON files into 11 JSON files, round-robin paths

In this example, we will generate 7 random JSON files. Next we will merge them and then split the merge into 11 files, round-robinning the paths across the 11 new JSON files:

$ randdict -B5 -D5 -c 7 -o data.json
$ ls
data000.json  data001.json  data002.json  data003.json  data004.json  data005.json  data006.json
$ jsonsam *.json - |split -xn r/11
$ ls
data000.json  data001.json  data002.json  data003.json  data004.json  data005.json  data006.json  x00  x01  x02  x03  x04  x05  x06  x07  x08  x09  x0a
$ ls x* |xargs -IF jsonsam F
Updated JSON file written to x00-norm.json
Updated JSON file written to x01-norm.json
Updated JSON file written to x02-norm.json
Updated JSON file written to x03-norm.json
Updated JSON file written to x04-norm.json
Updated JSON file written to x05-norm.json
Updated JSON file written to x06-norm.json
Updated JSON file written to x07-norm.json
Updated JSON file written to x08-norm.json
Updated JSON file written to x09-norm.json
Updated JSON file written to x0a-norm.json

The 11 x<num>-norm.json files now contain all the data in the original 7 files, but split round-robin across 11 files. Substituting split -xn l/11 for the split command would instead do a conventional partition rather than round-robin. Of course any intermediate mutations, additions and deletions could also be made in the same pipeline.

Comand Line Set Operations Examples

Set operations union |, intersection & and difference - can be performed on JSON files specified with positional arguments (lhs) and -F (rhs) options. The following examples will use the original simple.json file as the left-hand operand and simple_aux.json as the right-hand operand:

$ cat simple_aux.json
{
  "Bad": true,
  "beer": 3.1415,
  "NEW": "PATH",
  "young": {
    "Vodka": [
      27
    ],
    "bark": null,
    "guts": "but",
    "last": 66
  }
}

Difference/except simple.json - simple_aux.json

In the following example we perform the set difference of simple.json - simple_aux.json. Paths common to both simple.json and simple_aux.json are removed from simple.json.

Note the -I option can also be added for all operations to ignore leaf values in comparison. This is useful when one wishes to use an existing JSON file as a template for paths to remove regardless of value.

Also note removed list elements become null to preserve the ordinal of retained elements.

$ jsonsam simple.json -F simple_aux.json -e
Updated JSON file written to simple-norm.json
$ diff --unified=20 <(cat simple.json |jq --sort-keys '.') <(cat simple-norm.json |jq --sort-keys '.')
 {
-  "Bad": true,
   "beer": 3.14,
   "rots": "Our",
   "young": {
     "Vodka": [
-      27,
+      null,
       [
         "goes",
         2.718,
         {
           "Well": "water",
           "max": "plank"
         },
         "old"
       ]
-    ],
-    "bark": null,
-    "guts": "but",
-    "last": 66
+    ]
   }
 }

Click to expand additional set examples:

Intersection (simple.json & simple_aux.json)

Intersection (simple.json & simple_aux.json)

In this example we perform the intersection of the two JSON files. Note that -I is passed, so the leaf values are ignored allowing us to also intersect with a slightly less precise value of pi.

$ jsonsam simple.json -F simple_aux.json -iI
Updated JSON file written to simple-norm.json
$ cat simple-norm.json
{
  "Bad": true,
  "beer": 3.14,
  "young": {
    "Vodka": [
      27
    ],
    "bark": null,
    "guts": "but",
    "last": 66
  }
}

Union/merge (simple.json | simple_aux.json)

Union/merge (simple.json | simple_aux.json)

In this example we perform the union of the two JSON files. Note the conflicting beer key is overwritten by the right-hand side's value and the NEW item has been added.

$ jsonsam simple.json -F simple_aux.json -u
Updated JSON file written to simple-norm.json
$ cat simple-norm.json
{
  "Bad": true,
  "beer": 3.1415,
  "rots": "Our",
  "young": {
    "Vodka": [
      27,
      [
        "goes",
        2.718,
        {
          "Well": "water",
          "max": "plank"
        },
        "old"
      ]
    ],
    "bark": null,
    "guts": "but",
    "last": 66
  },
  "NEW": "PATH"
}

Random nested dictionary generator examples

A random dictionary generator is included in the package. It generates deeply nested dictionaries of realistic random data. Random nested lists are also supported. The full range of JSON types are supported. Each step in breadth and depth is uniquely randomized.

To generate a random dictionary of max breadth 3, max depth 4 consisting of 40% nested lists (compact form for brevity):

$ randdict -B 3 -D 4 -l 40 |jq -c '.'
{"game":{"yourself":{"difficult":"0181 Aguilar Parkways","always":{"before":530982.9961689024,"leader":{"kid":0.41913904357146525,"available":523048.62318369857},"effect":-609888.9269447384}},"believe":[[{"ok":0.24842658485754932,"kind":"91390 Williams Forges Apt. 824","charge":0.7294452894392176},"482 Bonnie Route",{"cup":null,"hotel":-292649,"bill":0.6952953662736593}],[0.47409833741964447,"093 Becker Meadow"]]},"standard":[{"former":null,"make":270989.99712404597,"pay":245327},{"couple":"Ward-Wright","say":{"state":["Miller LLC","Simpson, Cooper and Cole","Baird, Wilson and Barnes"],"though":"76018 Peterson Keys Suite 971","better":0.28193072232673766},"you":{"expect":null,"matter":{"hear":"49746 Johnson Mountain","artist":"Carr Group","interest":"2366 Miller Mission"}}},{"often":{"reason":"68465 Rosario Drive Apt. 223","early":{"actually":null,"memory":null,"theory":null},"nor":"White, Campbell and Thomas"},"hope":{"poor":2548096.1722068526,"others":372384,"from":["Vaughn PLC","Pena, Singh and Bryant",322924]}}],"truth":[{"state":{"two":{"response":0.08053812548862638,"success":null,"protect":"Long, Ross and Garcia"},"institution":null,"assume":1888192},"cut":[null,{"audience":null,"grow":"Alexander Inc","blood":"Williams, Smith and Hernandez"},1358976.5609615073],"himself":["39624 Guzman Mountains",-2605707.399841271]},["Wilson Group",0.2127797923458118,{"man":{"media":null,"member":725008.2758262581},"forget":1022493.1525128945,"guess":"Thomas Ltd"}]]}

API Example

The following example shows how to use the API to generate a random test dictionary, pick two overlapping subsets of the dictionary, then take the difference, union and intersection of these dictionaries:

>>> from jsonsam import DictSam
>>> from jsonsam import DictGen
>>> test_dict = DictGen().gen_fake_dict(breadth_rng=(2, 4), depth_rng=(3, 5), list_dist=(3, 1))
>>> dict_sam_orig = DictSam(test_dict)
>>> dict_sam_a = dict_sam_orig.random_dict_pick(60)
>>> dict_sam_b = dict_sam_orig.random_dict_pick(80)
>>> diff = dict_sam_b - dict_sam_a
>>> union = dict_sam_b | dict_sam_a
>>> intersect = dict_sam_b & dict_sam_a

NOTE: At the moment only JSON-serializable structures are supported by DictSam

Full API readthedocs coming soon...

Authors

See AUTHORS