Skip to content
Interactively explore complex data structures at the REPL with minimum keystrokes
Branch: master
Clone or download
Fetching latest commit…
Cannot retrieve the latest commit at this time.
Permalink
Type Name Latest commit message Commit time
Failed to load latest commit information.
.circleci
doc
resources
src/datawalk
test/datawalk
.gitignore
CHANGELOG.md
LICENSE
README.md
notes.md
project.clj

README.md

datawalk

https://github.com/eggsyntax/datawalk

Provides a simple interactive data explorer.

(datawalk.core/repl my-data-structure)

In Clojure and ClojureScript, we spend a lot of time exploring and manipulating data structures, sometimes enormous ones. I've whiled away innumerable hours at the REPL navigating through complex nested data structures, often heavily-nested maps with long namespaced keys. I prefer to minimize the amount of typing I have to do for repetitive tasks, and this is certainly one of those tasks.

So datawalk pretty-prints the top level of your data structure, and then lets you drill down repeatedly, often with just a single keystroke (the number representing the item's position in the printed list, as shown next to the item) plus enter.

Often the reason for drilling down at the REPL is to determine a path to pass to get-in, so datawalk will track and [p]rint the path from the original root to the level you're on (as long as it's a viable get-in path, which it's not in the case of sets and lists).

At any point in this process, you can [s]ave the current item or sa[v]e the path (from root to item) into a map of saved items which will be returned when you [q]uit. You can also e[x]it and return just the current item.

As mentioned, you can [p]rint the path to current data. You can also print [c]urrent data (useful when you want to see it untruncated), print the [m]ap of saved data, or print [h]elp to get a reminder of these commands. Capital C and M give you pretty-printed versions of current and saved data.

You can also move [b]ackward or [f]orward in time, or move [u]p a level, or jump back to the original [r]oot, or use [!] (function) to call an arbitrary single-arg │ function on the current data (jumping to the result).

datawalk tries to do one thing well: eliminate tedium and typing when navigating data structures. The learning curve should be trivial; please let me know if it's not and I'll try to change that. I hope you find it as useful as I have.

Demo (fully-interactive mode):

Demo

Installation:

Leiningen:

[datawalk "0.1.12"]

Details at https://clojars.org/datawalk

Usage summary:

Clojure:
  (datawalk.core/repl my-data-structure) to start datawalking.
Clojure and ClojureScript:
  (look-at my-data-structure), and then (w [command])

Commands:
  numbers: Enter any number to jump to the corresponding item
  q :quit              ; exit and return saved values if any
  x :exit-with-current ; exit & return just this value
  s :save-current      ; save to map of return values
  v :save-path         ; save path to map of return values
  b :backward          ; step backward in history
  f :forward           ; step forward in history
  r :root              ; jump back to original root
  ! :function          ; call a 1-arg fn on data, jump to result (clj only)
  u :up                ; step upward [provides list of referring entities]
  h :help              ; print help & return same ent
  p :print-path        ; print path from root to current item.
  m :print-saved-map   ; print the map of saved data
  c :print-full-cur    ; print the current data in full, without truncation

Fully-interactive version (Clojure-only)

Start the fully-interactive version with (datawalk.core/repl my-data-structure). From there, you just enter a single character followed by enter -- a number to drill down to a specific item in the displayed list, b to step back to where you just were, h for help, and so on. It should be extremely self- explanatory.

Semi-interactive version (Clojure and ClojureScript)

To use the semi-interactive version, let's assume you've referred the two key functions:

(require '[datawalk.core :as dw :refer [look-at w]])

Start by calling

user> (look-at my-data-structure)

This initializes datawalk's internal state, so that it's always tracking a piece of current data. From there, call (w *), where * represents any of the commands listed in the usage summary. So for example:

user> (w 2) ; drill down to item 2 in the numbered sequence
user> (w b) ; step backward in time
user> (w s) ; save the current data to the map of saved values
user> (w h) ; view the help (command summary)

After each step, the current state will be stored in datawalk atoms and can be freely accessed:

datawalk.datawalk/data stores the current data.

datawalk.datawalk/saved holds the map of stored values.

In the same ns, you can also access the-root, the-past, and the-future.

Additionally, datawalk's built-in printing tools are available:

user> (w m) prints the map of saved values.
user> (w p) prints the path from the root to the current data.

Demo (semi-interactive mode):

Demo

Why two versions?

I developed the semi-interactive version so I could use datawalk in ClojureScript. ClojureScript presents a challenge: while you can trivially get user input in Clojure with read-line, ClojureScript has no consistent way to do so (although some specific cljs repls like planck add that functionality themselves). Because of this issue, it's very difficult to create a true interactive program as described above.

I've gotten some interesting tips about it, and @mfikes kindly pointed to his abio library (https://github.com/abiocljs) which begins to address the problem. I'm hoping that at some point I'll manage to put together a truly agnostic solution (PRs welcome!).

But now that I've created the semi-interactive version as an interim solution, it turns out it has some real advantages. Some people may actually prefer it, since it keeps you in an ordinary REPL at all times. Its only downside is that you have to type a few more characters.

         Fully-interactive:                |   Semi-interactive:
                                           |
PROS:  - fewer keystrokes                  | - Runs anywhere
       - extremely fast                    | - Easy to do other stuff in the middle of a session
                                           | - You're always in a normal REPL
                                           |
CONS:  - Clojure-only                      | - Requires more keystrokes
       - Can't do other stuff in the       | - Not quite as fast to explore data
         middle of a session               |

Configuration

datawalk.print contains a config atom holding several entries which you can alter to change the behavior of datawalk. These configs all affect the representation of the current data structure, not the underlying behavior.

  • :max-items the top level of your data structure may be a sequence with hundreds or thousands of items; you probably don't want to print them all. Defaults to 30.

  • :max-line-length on each printed line, datawalk attempts to represent the entire contents of the data on that line. This can sometimes be enormously long. You may wish to tune it to match the width of your repl environment.

  • :max-key-length when displaying maps, the keys may be so long that they take up most of the space defined by *max-line-length*. This is most often true with namespaced keys. *max-key-length* limits the space taken up by the keys, in order to be sure to leave room for the values. When keys must be chopped, they're chopped from the right to try to capture the name rather than the namespace.

  • :debug-mode when this is truthy, datawalk will print the values of the current path, saved values, the-past, and the-future at each step.

datawalk.core also has a convenience function, set-line-length, which can be passed a single number. max-line-length will be set to this value, and max-key-length will be set to 20% of the value.

Note for Emacs users

Emacs accepts user input (via read-line) in the minibuffer -- this can be confusing when you first encounter it; the REPL buffer looks like it's hung. Just look down at the minibuffer and you'll see you can enter a command there. Note that it's important to q)uit the datawalk session, or your REPL buffer may be left in a problematic state.

License

Copyright © 2017 Jesse Davis (aka Egg Syntax)

Distributed under the Eclipse Public License either version 1.0 or (at your option) any later version.

You can’t perform that action at this time.