SimplyStats Thanksgiving Puzzle -

I owe a lot to Jeff Leek and Roger Peng for their great Coursera courses, in which I learned to program in R.

They (along with Rafa Irizarry) run the Simply Statistics (simplystatistics.org) blog, which I highly reccomend. They posted a Thanksgiving puzzle in which a data.frame needs to be converted from one form to another, spelling out ‘thanksgiving’.

http://simplystatistics.org/2015/11/25/a-thanksgiving-dplyr-rubiks-cube-puzzle-for-you/ (simplystatistics.org)

The puzzle: convert this

	col3 col2 col4 col1
	h a t t
	i v i g
	k s g n
	n g n i

view raw (gist.github.com) puzzle_nov25_2015.delim (gist.github.com) hosted with ❤ by GitHub (github.com)

into this

	col1 col3 col2
	t h a
	n k s
	g i v
	i n g

view raw (gist.github.com) tday.delim (gist.github.com) hosted with ❤ by GitHub (github.com)

My solution, which uses Rubik’s Cube rotations of rows and columns (and dplyr of course):

	## transform the input to the output using rubiks-cube transformations and dplyr
	require(dplyr)

	## provided input
	input <- data.frame(matrix(c("h","a","t","t","i","v","i","g","k","s","g","n","n","g","n","i"),
	4,
	4,
	byrow=TRUE,
	dimnames=list(NULL,paste0("col",1:4))),
	stringsAsFactors=FALSE)

	## requested output
	output <- data.frame(matrix(c("t","h","a","n","k","s","g","i","v","i","n","g"),
	4,
	3,
	byrow=TRUE,
	dimnames=list(NULL,paste0("col",1:3))),
	stringsAsFactors=FALSE)

	## construct the column rotating function using dplyr commands
	## taking care of wrap-around
	## allows forwards or backwards rotation through + or - operator
	rotate_col <- function(df, colnum) {
	if(colnum > 0) {
	## rotate the cube 1->4 by 1 step (row3->row4, row2->row3, etc...)
	## row4->row1
	rotdfcol <- df %>% slice(c(2,3,4,1)) %>% select(colnum)
	} else {
	## rotate the cube 4->1 by 1 step (row4->row3, row3->row2, etc...)
	## row1->row4
	colnum <- -colnum
	rotdfcol <- df %>% slice(c(4,1,2,3)) %>% select(colnum)
	}
	newdf <- df
	newdf[, colnum] <- rotdfcol
	return(newdf)
	}

	## construct the row rotating function using dplyr commands
	## using the same conventions as rotate_col
	rotate_row <- function(df, rownum) {
	if (rownum > 0) {
	rotdfrow <- df %>% select(c(4,1,2,3)) %>% slice(rownum)
	} else {
	rownum <- -rownum
	rotdfrow <- df %>% select(c(2,3,4,1)) %>% slice(rownum)
	}
	newdf <- df
	newdf[rownum, ] <- rotdfrow

	return(newdf)
	}

	## perform rubiks-cube operations to achieve the desired result, utilising row 4 as free spaces
	## in no way guaranteed to be the fastest method, just the first one that worked
	input %>%
	rotate_row(+1) %>%
	rotate_col(+4) %>% rotate_col(+4) %>%
	rotate_col(-1) %>% rotate_row(+4) %>% rotate_col(-1) %>% rotate_row(-4) %>% rotate_col(+1) %>% rotate_col(+1) %>%
	rotate_col(-2) %>% rotate_row(+4) %>% rotate_col(-2) %>% rotate_row(-4) %>% rotate_col(+2) %>% rotate_col(+2) %>%
	rotate_col(-3) %>% rotate_row(-4) %>% rotate_col(-3) %>% rotate_row(+4) %>% rotate_col(+3) %>% rotate_col(+3) %>%
	rotate_row(+4) %>% rotate_col(-1) %>% rotate_row(-4) %>% rotate_col(+1) %>%
	rotate_col(-2) %>% rotate_row(+4) %>% rotate_col(+2) %>%
	rotate_row(+4) %>% rotate_col(-3) %>% rotate_row(-4) %>% rotate_col(+3) %>%
	rotate_col(-4) %>% rotate_row(+4) %>% rotate_col(+4) %>%
	slice(c(1,2,3)) -> solved ## still a 3x4 matrix

	## re-shape the solution into a 4x3 matrix
	final <- solved %>%
	t %>%
	c %>%
	matrix(4,
	3,
	byrow=TRUE,
	dimnames=list(NULL,paste0("col",1:3))) %>%
	data.frame(stringsAsFactors=FALSE)

	## check the solution is correct
	identical(final, output) # TRUE

view raw (gist.github.com) simplystats_thanksgiving.R (gist.github.com) hosted with ❤ by GitHub (github.com)

Suggestions on how I could have done this differently (or automated solutions) most welcome!

rstats

See also