Tips for golfing in 2D languages

Question

What general tips do you have for golfing in 2 dimensional programming languages? I'm looking for tips which can be applied to code-golf problems and are specific to 2D programming languages, but not specific to any one language (Responses like "remove comments" and "use the M operation" are not answers).

Please post one tip per answer.

Answer 1

Avoid horizontal gaps

Often, code will leave largs gaps of whitespace to the left hand side of the program, like so.

abc
  d
  e

This adds 4 bytes, when this could be avoided by left aligning.

cde
b
a

If you need to use large gaps of whitespace, try to make them vertical, instead of horizontal.

########
#      #
#      #
#      #

vs

####
#
#
#
#
#
#
####

Answer 2

Carriage Returns Are Bytes Too

The less 2D you can make it, the better. A carriage return is another no-op. Without ignoring the tips from @ATaco and @ASCII-only, try and keep the Y dimension as small as possible.

This

###
####
########

is better than

###
###
###
##
#
#
#
#

Answer 3

Use one dimension when possible

Typically, simpler programs can be written on a single line. For example, the classic cat program could be:

>iv
^o<

But one could abuse the wrapping behavior and make this:

io

Or, in languages without such wrapping behavior:

> ?oi<

(Assuming ? doesn't pop.) In the case of a non-wrapping language, an explicit loop is often better.

With jump commands

In 2D languages with jump and conditional jump commands, a program could look like this:

abc >de?v;
    ^hgf<

This could also be:

abc de?!;hgf04&

(if ! is a trampoline, and & is jump to position)

Answer 4

DRY (Don't Repeat Yourself)

While abstracting with functions is usually longer in Code Golf, it can really help for 2D languages. Try to rework your code so it can re-use the same snippet, entering/exiting it with two different branches of execution.

Answer 5

Interleave paths

Usually in a 2D language there is an IP that moves according to direction commands. Since spaces are wasted bytes, it is almost always more efficient to rearrange the program so it moves closer to the left as often as possible, saving the need for unnecessary padding spaces.

Answer 6

Use mirrors

Mirrors can sometimes be used in two paths at the same time (each path bounces off one side of the mirror). This may not seem to help, but it may allow you to rearrange your program, or if you have a lot if direction changes they may be able to be replaced with fewer mirrors.

Answer 7

Memorize idioms

Here are a few "idioms" that do certain things, depending on the nature of the language.

Pseudo-linear code

If dynamic code generation is ever required, it may be of use to use the pseudo-linear code model:

v
\"line 1"
\"line 2"
.
.
\"line N"

Assuming \ and v mean what they usually do.

Infinite loop

In almost all 2D languages, >< is an infinite, unbreakable loop. If, for some reason, you need to do this, this is the best way, despite how nice this might look:

>v
^<

In fact, if you make your code a 1-liner, you could just use ^ or v, as such:

i?vo;

This v will send the IP to itself, wrapping around. You may still be able to use this approach in any instance where a directional command points to a series of (relative) no-ops.

Quine framework

Usually, languages with a string/quote framework can have a quine like this:

<quote><generate "><output stack><terminate>

For ><>, this would look like:

":1-r>o<#

Except this one exits with an error as termination. It is probably the shortest ><> quine, or, at least, the shortest one that I have found.