A CLI utility written in pure Rust for stream converting JSON objects to a series of CSV values, from stdin and/or file(s) to stdout or a file.

• Install rust
  1. Install from cargo
     • cargo install jaesve
  2. Install from the repo
     • Run rustup default stable
     • Run git clone https://github.com/bazaah/jaesve.git; cd jaesve
     • Run cargo run --release
     • The binary can be found in target/release/

Jaesve comes with a CLI, courtesy of clap.rs. You can type -h or --help to see the available settings, or browse a complete listing below.

• -h --help Displays help (use --help for more detailed messages)
• -V --version Displays version information
• -a --append Append to output file, instead of overwriting
• -q --quiet Silences error messages
• -v Verbosity of debug information
  • Max: 3

• -o --output Set output file to write
  • Default: stdout
• -l --line Set stdin to read a JSON doc from each line, and the line to start processing from
  • Default: 0
• -E --regex Set a regex to filter output
• -c --column Sets field to match regex on
  • Possible: ident, jptr, type, value, jmes
• -f --format A dot '.' separated list of fields describing how output is formatted
  • Default: ident.jptr.type.value
  • Possible: ident, jptr, type, value, jmes, all
• -d --delim Sets delimiter between output fields
  • Default: ,
• -g --guard Set field quote character
  • Default: "

• config Configure various program intrinsics
  • These parameters are unstable and may change in the future
  • config --help for the current list
• completions Autocompletion script generator
  • completions [FILE] -- <SHELL>
    • If no FILE is present, defaults to stdout
    • Possible SHELLs: bash, zsh, fish

• A space separated list of valid file paths, with a - representing stdin. If you wish to add flags and options after this Arg, you must end its values with a :.
  • For example:
    • jaesve --quiet --append -o output.csv input1 input2 - input4 // This is happy
    • jaesve input1 input2 - input4 --quiet --append -o output.csv // This is sad
    • jaesve input1 input2 - input4 : --quiet --append -o output.csv // This is happy

A subset of the above flags and options can be set via environment variables or config file(s). Listed below:

• debug (-v)
• quiet
• append
• line
• delim
• guard
• format
• buf_in
• buf_out
• linereader_eol
• factor

They expect the same input kinds as the CLI variants, with flags (i.e quiet) taking common bool representations -- e.g true, No, 1, etc.

When given a variable from multiple sources the program will prioritize in this order: (highest to lowest)

1. CLI
2. Environment
3. $HOME/jaesve.conf
4. $CONFIG/jaesve.conf
5. /etc/jaesve.conf (if on *nix)
6. Default (if any)

The program will check for variables with the format JAESVE_<VAR_NAME>.

Config files are disabled by default, if you wish to use them add --features=config-file to your cargo install/build. The expect format is TOML and an example file is listed below:

quiet = false
delim = ","

# Note this corresponds to the 'config' CLI subcommand
[config]
factor = "K"

In preliminary tests it parsed 2G of JSON in 3 minutes.

Jaesve is written to minimize memory usage. It uses a stream based approach to parsing JSON, and attempts to unroll nested objects. Its maximum memory footprint can be described as follows:

• sizeof largest object/array +
• sizeof combined elements NOT including any object/array from doc start to largest object/array +
• program overhead

TLDR: the more deeply nested and larger the objects/arrays, the larger the memory footprint.

For a simple example, let's use the following JSON:

//sample.json
{
  "aliceblue": "#f0f8ff",
  "antiquewhite": "#faebd7",
  "azure": "#f0ffff",
  "beige": "#f5f5dc",
  "black": "#000000",
  "blanchedalmond": "#ffebcd",
  "gradient": {
    "blues": ["#0000f0", "#0000f1", "#0000f2"],
    "green": "#00ff00"
  }
}

Running jaesve sample.json prints out

"1","/gradient/blues/0","String","#0000f0"
"1","/gradient/blues/1","String","#0000f1"
"1","/gradient/blues/2","String","#0000f2"
"1","/gradient/green","String","#00ff00"
"1","/gradient/blues","Array",""
"1","/aliceblue","String","#f0f8ff"
"1","/antiquewhite","String","#faebd7"
"1","/azure","String","#f0ffff"
"1","/beige","String","#f5f5dc"
"1","/black","String","#000000"
"1","/blanchedalmond","String","#ffebcd"
"1","/gradient","Object",""

Where:

• "INTEGER" is which input source the value came from
• "/.../..." is the json pointer of that record
• "JSON TYPE" is the record type
• VALUE is the value associated with that record, if it is an endpoint i.e not a Object or Array

Jaesve prints any errors to stderr unless --quiet is set with escalating information on -v, -vv and -vvv.

It will error for the following conditions:

1. JSON object is malformed / contains invalid unicode points
2. Could not read from input file(s)
3. Could not create output file (It will not create directories)
4. Could not write to output

The above errors are expected to be handled by the user, if other errors occur they are considered bugs and I'd appreciate it if you'd open an issue with a description of what went wrong and the error message you received.