michaelrauh / haskell_fold Goto Github PK

The functions tend to use concrete types which is under constraining

Can a box be made from a single word?

In dimension boxes can only be combined with other in dimension boxes, but this is not reflected in the type system. It would be better to wrap both boxes with the combinable so that there is no opportunity for an in dimension box to attempt to combine with a next dimension box.

There are two obvious fixes for this:

1. Hash things when possible (this should give benefits in comparison speeds too
2. Add the ability to build things incrementally by adding chunks of text one at a time instead of all at once. This will allow it to get through bottlenecks and GC out the intermediate data.

tracking of lines is going to require zipping orthotopes together

In the first chapter of A Princess of Mars (from "To the Reader of this Work" to "upon the floor." there is a 3x3 that looks like:

to him I
that he was
I wished to

That is not being found.

This would be a more elegant solution. Currently it takes a bigram, but tracking should work without bootstrapping at this point.

Either override eq and ord to recognize that transposes are equivalent, or skip any dimension searches that are the same dimensions but rotated 90 degrees.

Here is known missing data:

1. Lines
2. Columns

Are there missing pieces of data? How will center overlap be validated?

Corners overlap gets rid of the trivial case but there are evidently other cases that produce bloat in the output. The easiest fix for this is to make sure that all words are unique but that may be overkill

The new design requires a concatenated string representing a whole line that points to a column word

1. Attempt to take out the point type
2. Attempt to strip the type down to a plain list of list