rpick is a command line tool that helps you to pick items from a list, using configurable algorithms.

An example use case for this is picking a restaurant. You might want to generally go to restaurants you haven't visited in a while, but you also might not want to use a strict least recently used model and spice things up with some element of chance, with restaurants you've least recently visited getting a boost in their chances.

Install

You can install rpick with an Arch AUR package:

$ git clone https://aur.archlinux.org/rpick.git
$ cd rpick
$ makepkg -sirc

Or in Fedora:

$ sudo dnf install rpick

Or in Gentoo:

$ sudo emerge app-misc/rpick

Or in MacOS, assuming you have installed brew:

$ brew install rust
# Be sure to read the output of this command and adjust your PATH as instructed.
$ cargo install rpick

If you are not using one of the systems described above, rpick is also available on crates.io. You can install rpick this way by first installing Rust, and then using Rust's cargo tool to install rpick:

# Be sure to read the output of this command and adjust your PATH as instructed.
$ cargo install rpick

Quick start

rpick keeps its state in a YAML file in your home config directory called rpick.yml. For now, users must create this file by hand, and rpick will manage it from there. To get started with some examples, create ~/.config/rpick.yml in Linux, ~/Library/Preferences/rpick.yml on MacOS, or C:\Users\Alice\AppData\Roaming\rpick.yml on Windows, like this:

---
prs:
  model: even
  choices:
    - paper
    - rock
    - scissors
restaurant:
  model: gaussian
  choices:
    - Spirits
    - Lucky 32
    - Centro
    - Sitti
    - Cookout

Then you can ask rpick to pick a game of paper rock scissors for you:

$ rpick prs
Choice is scissors. Accept? (Y/n)

Note that it would be bad to use the Gaussian model for paper rock scissors, because you have a statistical advantage of guessing that model's results. Let's take a look at the Gaussian model:

$ rpick restaurant
Choice is Lucky 32. Accept? (Y/n)

If you say yes, it will rewrite the yaml file like this since we used the Gaussian model:

---
prs:
  model: even
  choices:
    - paper
    - rock
    - scissors
restaurant:
  model: gaussian
  stddev_scaling_factor: 3.0
  choices:
    - Spirits
    - Centro
    - Sitti
    - Cookout
    - Lucky 32

Note that we passed prs and then restaurant as arguments when we called rpick - this told rpick to look for those objects in rpick.yml to find out which models to use and which choices were available. This parameter is required, but its possible values are defined by you in your config file.

The model field in the config file defines which mathematical model to use to pick from the given choices. See the Models section below for more information about which models are available and how you can configure them.

It added one setting to your restaurant object that wasn't there originally: stddev_scaling_factor. You can read more about this setting in the Gaussian model documentation below.

Parameters

The CLI accepts a few parameters:

• -c/--config: This can be used to specify an alternate path for a config file for rpick to use. You can also set the RPICK_CONFIG environment variable.

• -h/--help: Print help text.

• -v/--verbose: Print more information about the pick.

• -V/--version: Print the rpick version.

Models

rpick is capable of a few different algorithms for picking choices: even, gaussian, inventory, lottery, lru, and weighted.

Even

The even distribution model is the simplest available choice model. It will give an even chance to each item in the list of strings to be chosen. It requires two keys:

• model: This must be set to the string "even", in order to select this model.
• choices: This is a list of strings that are the options for the model to choose from.

Example:

convertible_top:
  model: even
  choices:
    - up
    - down

You might want to consult the weather before using rpick for this use case…

Gaussian

To understand the Gaussian distribution model, we need to first study this XKCD comic:

The gaussian distribution model is more complex. It uses the Gaussian distribution to prefer choices that have been less recently chosen. Things near the top of the list of choices have the highest probability of being chosen, while things at the end of the list have the lowest chance. Once an item has been picked and the user has accepted the choice, the list is saved to disk with the picked item moved to the end of the list. This model accepts three keys:

• model: This must be set to the string "gaussian", in order to select this model.
• stddev_scaling_factor is used to derive the standard deviation; the standard deviation is the length of the list of choices, divided by this scaling factor. Thus, a larger scaling factor will result in a stronger preference for items near the top of the list, and a smaller scaling factor will result in a more even distribution among the choices. Note that the smaller the scaling factor is, the longer rpick will take to make a decision, on average. The default is 3.0, which is chosen because it places the last item on the list at three standard deviations, giving it a 0.03% chance of being chosen. This key is optional, and defaults to 3.0.
• choices: This is a list of strings that are the options for the model to choose from.

Example:

album:
  model: gaussian
  stddev_scaling_factor: 5.0
  choices:
    - Fountains of Wayne/Fountains Of Wayne
    - Beck/Odelay
    - "Townes Van Zandt/High, Low and In Between"
    - Tori Amos/From The Choirgirl Hotel
    - Zao/Parade Of Chaos

To show a visual representation of the relative frequencies of this model, here is a graph that was generated by calling the Gaussian model 10,000 times with the choices set to the numbers 0 to 99 sequentially and a default stddev_scaling_factor of 3.0:

Inventory

The inventory distribution model is a dynamic version of the weighted model. Each of the choices has a certain number of lottery tickets that influence how likely they are to be picked that round. Once an item is picked, it loses one ticket, i.e., dropping the inventory of that particular item by one. It accepts two keys:

• model: This must be set to the string "inventory", in order to select this model.
• choices: This must be a list of objects. Each object accepts two keys:
  • name: This is required, and is the name of the choice.
  • tickets: The current number of lottery tickets that this choice has. This is optional, an integer, and defaults to 1.

Example:

tea:
  model: inventory
  choices:
    - name: "Tea… Earl Grey… Hot"
      tickets: 15
    - name: Black
      tickets: 2

Lottery

The lottery distribution model is also a dynamic version of the weighted model. Each of the choices has a certain number of lottery tickets that influence how likely they are to be picked that round. Once an item is picked, it gets its lottery tickets reset to reset tickets and every choice that wasn't picked gains more lottery tickets. It accepts two keys:

• model: This must be set to the string "lottery", in order to select this model.
• choices: This must be a list of objects. Each object accepts four keys:
  • name: This is required, and is the name of the choice.
  • reset: How many tickets the choice is reset to when picked. Defaults to 0.
  • tickets: The current number of lottery tickets that this choice has. This is optional, an integer, and defaults to 1.
  • weight: This is an integer expressing how many lottery tickets are given to this choice when it is not chosen. You can use this to influence how often this item gets favored relative to the other choices. It is optional, and defaults to 1.

Example:

activity:
  model: lottery
  choices:
    - name: exercise
    - name: read documentation
    - name: watch tv
      weight: 1000

LRU

The lru (Least Recently Used) model is a simple ordered FIFO (First In, First Out) list. The least recently chosen item is the first item in the list. On each pick, rpick simply picks the first item in the list that the user says yes to, and then moves that item to the end of the list. It accepts two keys:

• model: This must be set to the string "lru", in order to select this model.
• choices: This must be a list of strings. The least recently chosen item should be at the beginning of the list.

Example:

code:
  model: lru
  choices:
    - Add a feature
    - Fix a bug
    - Write some docs

Weighted

The weighted distribution model is a more general version of the even model that allows you to express different weights for each of the choices. It accepts two keys:

• model: This must be set to the string "weighted", in order to select this model.
• choices: This must be a list of objects. Each object accepts two keys:
  • name: This is required, and is the name of the choice.
  • weight: This is an integer expressing the weight for the choice. It is optional, and defaults to 1.

Example:

cereal:
  model: weighted
  choices:
    - name: generic bran flakes
    - name: cracklin oat bran
      weight: 1000

Changelog

See the Changelog.

Contribute

If you would like to contribute to rpick, send me a patch!

There is a Makefile that is handy for development, if you have podman on your system. It's default target is a help menu that describes the available targets.

# Note that root is not required.
$ make check

The podman development environment looks for an rpick config at config.yml.