iogii ⸱ Docs ⸱ Source ⸱ Quick Ref ⸱ Online Interpreter

Source

To run iogii save the "1 File" from the latest release (below) as iogii and chmod +x it, then run iogii <filename.iog> (it is standalone ruby code with no dependencies).

You can also download "All Source" and run cd src; ruby main.rb <filename.iog>. This includes the original source files, all tests, and site generation code.

The current version is compatible with Ruby 2.7.2 - 3.1.2. And if you use Haskell code generation (with the -hs flag), it is compatible with GHC 8.10.7 - 9.2.2. These are only the versions I have tested with in Ruby + Haskell, it is likely to work in other versions as well.

I am not using github due to them not respecting licenses by training LLMs on code without crediting said code when used in a response.

0.3 (Beta) - March 18, 2025

• 1 File
• All Source
• Website Snapshot

Apologies for the large number of major changes. I have no further plans for large changes and should just be making minor changes and fixing bugs/documentation in the future.

Changes:

• Remove block syntax, now ops that create loops act like regular values syntactically
  • Removes need for ! and nested rotation trick, instead just place values that needed to be to the left of the arg to the left of the loop op.
  • Allow >, to match previous loop creation ops skipping initial match
• > in main function overrides implicit values like how @ did.
• Additional implicit values beyond the initial input use the $ of where they are used (they are also placed at beginning of unrotated program).
• Auto input is removed from the stack if empty and implicits used instead.
• Add set next var (=).
• Remove [, ], and @ ops, use new = instead.
• mdup / mpeek cannot match > (this allows them to always be skipped when matching >, use set next var instead if you needed that).
• Rename assignment from ~ to set and add let syntax (alias for set ... del).
• Change broadcasting when there is excess rank.
• Better handling of ops with two generic types, e.g. sortBy.
• Fix an assortment of bugs (please report any new ones!).
• Support negation in data format, e.g. 1,2~,3 = [1,-2,3]
• Use the system's character encoding rather than bytes. You can force a string literal to still use bytes with ""b.
• - cannot subtract char from char (use ^ now), this is to uphold an invariant that makes circular type inference sound.
• Change charClass op to be a unary function that returns all chars in given character class.
• iterate takes a list to prepend instead of a single value
• empty auto input is nil instead of empty string
• Empty type defaults to Int when used as a scalar (before errored).
• You can do a vectorized op with a list of empty type (nil), this would usually be useless since result will always be an empty list (before errored).
• keep promotes first via repeat (to easily generate a fixed number of scalars for example).
• chunkWhen returns rest of first list when second list runs out
• Raw mode for stdin does not break input into lines (also created with , now).
  • Add lines (p) low rank overload to approximate old behavior.
• Add peek, mpeek, consDefault, rightJustify, dropUntilAfterSubstring, keepUntilSubstring, getSubstring, and init ops
• Remove or (use max or ifElse instead)
• Renaming of several ops, the purpose of which is to provide better mnemonics and make some ops more generic.
  • = → q (equal)
  • s → & (sortBy)
  • q → s (len)
  • # → _ (sum)
  • P → . (prod)
  • | → L (abs) (low rank overload)
  • _ → ~ (negate / read)
  • | → . (ord)
  • P → _ (words)
  • A → (, (min) (now generic and not a low rank overload).
  • X → x (max) (now generic and not a low rank overload).
  • x → \ (setDiff)
  • & → # (reshape)
  • \ → | (chunkWhen)
  • W → # (fromBase) (W not valid overload with new behavior of a b types).

0.2 (Alpha) - Dec 11, 2024

• 1 File
• All Source
• Website Snapshot

Changes:

• Clarified reshape behavior, explained in new Ops page.
• Fix bug with , following token that was split up, e.g. nJ,
• Add Graphviz visualizations to online interpreter.
• Support Haskell code generation (~50x speedup + no stack overflow limit)
• Remove pp
• show now prints like Haskell would
• Rename nuke to del
• Remove regex match
• Add charClass op
• Output chars mod 256 rather than error if out of range
• expand does not return the initial value now (thanks UnrelatedString for the suggestion).
• Change transpose to insert default values so that truncation is never needed.
• Support input from command line args after --.

0.1 (Alpha) - Nov 7, 2024

• 1 File
• All Source
• Website Snapshot

Changes:

• Add more string ops
  • charRange ^
  • uppercase U
  • lowercase L
  • parts P (split on whitespace)
  • chr |, (convert int to char)
  • match + to match regexes.
• Allow unterminated strings.
• Allow implicit args.
• Add auto parsing of input.
• @ also sets input if none present.
• Make pow ^ work more consistently for negative numbers and not overflow for large numbers.
• Remove ops: nfoldProd p, and y
• Add ops: pad p, index ., get g
• Rename ops:
  • chunkWhen . → \
  • transpose \ → p (pivot). So that it can be unvectorized and also provide nice overload for prod.
  • digits \ → p (it's like transposing a string)
  • last z → l
  • len l → q (quantity)
  • equal q → =
  • assignment = → ~
  • generate g → K (kCopy), no longer generic
  • if → y (yes no)
  • sum _ → #
  • negate / read ~ → _
  • pow10 Z → H (hyper)
  • sqrt H → u (unsquare)
  • toBase # → D (digitize)
  • fromBase #, → W baseFrom (base was). Note that arg order is also swapped - necessary to have vectorized overload.

0.0 (Alpha) - Oct 29, 2024

Initial release. Uses Ruby 3.1.2.

• 1 File
• All Source

SHA256 sum of 1 File: d325338fab8c0d44a94a2dd85c9deda61d32fe20521e4c125c99d142e37e5a5c