Skip to content
/ lpl Public
forked from SOF3/lpl

Command-line plotting from streaming input sources

License

Notifications You must be signed in to change notification settings

rustbunker/lpl

 
 

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

lpl

A command-line utility to plot charts from line-based inputs.

Sinusoidal plot example

Features

  • Multiple data sources in different formats as polling or streaming inputs:
    • JSON (inotify + periodic reload)
    • JSONLines (streaming input)
    • CSV (inotify + periodic reload)
    • CSV (streaming input)
  • Interactive scrolling
  • Series hiding/color selection

Example usage

A few simple use cases using Bash:

Device thermometers

Linux exposes the device thermometers under /sys/class/thermal.

For example, on my laptop:

$ cat /sys/class/thermal/*/temp
47000
20000
39050
45000
47050
40000
47000

Therefore, we can pass each file as a separate --csv-poll argument to lpl:

$ lpl $(
>     for file in /sys/class/thermal/*/temp; do
>          echo "--csv-poll $(basename $(dirname $file))=$file"
>     done
> )

Rolling deployment

When performing a rolling update with a Kubernetes Deployment, it scales down the old ReplicaSet and scales up the new ReplicaSet gradually. We can monitor this progress by watching the ReplicaSet replica count using <() process substitution with kubectl get --watch and some transformation with jq:

lpl --json <(kubectl get replicaset --watch -ojson | \
        jq --unbuffered -c '{
            (.metadata.name + " total"): .status.replicas,
            (.metadata.name + " ready"): .status.readyReplicas,
        }')

Currently lpl only supports JSONLines, so make sure to pass -c to print one JSON object per line. Also use --unbuffered to ensure jq can emit new events immediately.

Reference

Input sources

JSON

JSON comes in two modes — streaming and polling.

Streaming JSON is specified by --json PATH, where PATH is a special file (typically from <() process substitution) that reads one JSON object per line. A JSON object may contain arbitrary fields, but only top-level fields with a single numeric value are processed, where the numeric value is added to the time series named with the corresponding key, at the time the line is read. Pipe your input to | jq -c --unbuffered . if it was not already one line per object.

Polling JSON, on the other hand, is a (usually regular) file that contains a single JSON object (either compact or formatted), specified by --json-poll PATH. Similar to streaming JSON, only top-level fields with a single numeric value are processed. The file is reloaded when it is changed (where supported by inotify), or every --poll-period seconds.

CSV

Similar to JSON, CSV also supports streaming and polling modes through --csv PATH and --csv-poll [HEADER=]PATH.

CSV files are separated with , by default, but this may be customized with --csv-poll-delimiter.

If the argument contains =, the part before = is treated as the CSV header. Otherwise, the first line read from PATH (initial first line for streaming, first line from every reload for polling) is treated as the CSV header.

Only numeric values that can be parsed as f64 are processed; other values are silently ignored. For polling mode, if there are multiple files in a single poll, only the first numeric value is processed.

Interactive CLI

lpl provides an interactive TUI to browse the data plot.

Type ? for help. Type q to quit.

Installation

Compile from source

Use cargo install:

cargo install lpl

Development builds

Download development builds built on GitHub CI:

About

Command-line plotting from streaming input sources

Resources

License

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published

Languages

  • Rust 100.0%