___           ___
     /\  \         /\  \
    /::\  \       /::\  \
   /:/\:\  \     /:/\:\  \
  /::\~\:\  \   /::\~\:\  \
 /:/\:\ \:\__\ /:/\:\ \:\__\
 \/__\:\/:/  / \/__\:\/:/  /
      \::/  /       \::/  /
       \/__/         \/__/

A fast, single-namespace, no-dependency pretty-printer for data (not code).

Supports Clojure, ClojureScript, and Babashka.

Features

• Fast¹ (~14x-10x faster than fipp.edn/pprint and ~85x–16x faster than clojure.pprint/pprint at Fipp's benchmark)
• Small (under ~500 lines of code, not counting comments)
• Zero dependencies
• Single namespace; either use as a dependency or vendor into your codebase
• Output similar to clojure.pprint/pprint
• Allocates conservatively (~28x fewer allocations than clojure.pprint/pprint, ~12x fewer than fipp.edn/pprint)

Non-goals

• Formatting code
• API for large-scale output customization; to customize, vendor and change whatever you want
• cl-format
• Syntax coloring (á là Puget)

Use

Either:

• Copy src/me/flowthing/pp.cljc into your codebase and rename the namespace to avoid conflicts, or:

• Pull in via Maven or Git:

  ;; Maven coordinates
  me.flowthing/pp {:mvn/version "2024-11-13.77"}
  
  ;; Git coordinates
  io.github.eerohele/pp {:git/tag "2024-11-13.77" :git/sha "13fba3a"}

Then:

user=> (require '[me.flowthing.pp :as pp])
nil
user=> (pp/pprint {:a 1 :b 2 :c 3 :d 4})
{:a 1, :b 2, :c 3, :d 4}
nil
user=> (pp/pprint {:a 1 :b 2 :c 3 :d 4} {:max-width 10})
{:a 1,
 :b 2,
 :c 3,
 :d 4}
nil

API

user=> (require '[me.flowthing.pp :as pp])
nil
user=> (clojure.repl/doc pp/pprint)
...

The API is stable and there will be no breaking changes. Everything except for me.flowthing.pp/pprint is internal and subject to change.

Differences to clojure.pprint

Even though pp is not meant for formatting code, having it format the source of every var in the clojure.core namespace and comparing the output to that of clojure.pprint/pprint is a good exercise because clojure.core has a large variety of data structures and nesting levels.

In that exercise, every difference between the outputs of me.flowthing.pp/pprint and clojure.pprint/pprint is one where clojure.pprint/pprint doesn't make full use of the 72 character line width (default for both clojure.pprint/pprint and pp) even though it could.

Also, unlike clojure.pprint/pprint, pp prints records like pr does:

user=> (defrecord R [x])
user.R
nil
user=> (prn (->R 1))
#user.R{:x 1}
nil
user=> (pp/pprint (->R 1))
#user.R{:x 1}
nil
user=> (clojure.pprint/pprint (->R 1))
{:x 1}
nil

Furthermore, clojure.pprint/pprint prints Java arrays with commas, pp prints them without:

user=> (clojure.pprint/pprint (char-array [\c \p \p]))
[\c, \p, \p]
user=> (me.flowthing.pp/pprint (char-array [\p \p]))
[\p \p]

In addition, there are one or two other minor, insignificant differences in where clojure.pprint/pprint and pp insert line breaks. If you spot these and they bother you, file an issue.

Differences to Fipp

• pp uses print-method for pretty much everything except Clojure's built-in collection types. This means pp, unlike Fipp, prints things like time-literals the same way as clojure.pprint/pprint.
• Fipp prints (fipp.edn/pprint '@foo) as (clojure.core/deref foo); pp, like clojure.pprint/pprint, prints it as @foo. The same with quote/', var/#', and unquote/~.

Notes on ClojureScript support

• ClojureScript doesn't support *print-dup*, so neither does pp.

Applications

• Tutkain
• Tab
• Lingo
• Bosquet
• Metabase
• Penpot
• Fireworks
• Kushi
• Yetti

Acknowledgements

The algorithm pp uses is based on the ideas in Pretty-Printing, Converting List to Linear Structure by Ira Goldstein (Artificial Intelligence, Memo No. 279 in Massachusetts Institute of Technology A.I. Laboratory, February 1973).

Footnotes

1. I have only benchmarked the Clojure implementation. Cursory testing with simple-benchmark shows a ~6x improvement over cljs.pprint/pprint, but grain of salt and all that. ↩