as_plate should probably be able to take the following:

data.frame

• either tidy = TRUE (requires - minimally - numeric columns x and y. The rest of the columns will be considered values and the column names the keys
• tidy = FALSE - a matrix of values, with x and y 'as is', and a name supplied as a separate arg

matrix/array

• almost exactly the same as data.frame for matrix
• for array, can either supply names equal to the extent of the 3rd dimension, or use the dimnames of third dimension if not null. Will override dimnames if names is supplied
• Must not accept >3D arrays

vector

• Will form column of wells. Requires names. Can be transposed after with t()

well

• simply list(well)

list

• largely will lapply any of the above

While this might seem easy, you have to think about:

1. Not all wells may contain the same keys in content
2. Not all values for a given key may be of the same type

Might be nice to be able to access a well/a set of wells by doing something like

my_plate[1, ] # return row 1 of wells
my_plate[1,2] # return well in position row 1, col 2
my_plate[, 2] # return col 2 of wells

What should happen? should it coerce to the lowest common denominator? error? warn?

There should probably be some canonical ordering of wells in a list. This makes for easier testing, as well as more reliable behavior. I'd imagine ordering by row, then col is best.

Making wells should be dead simple, I think.

Some things that should be possible:

# Create a row of wells
well(1:10, 1)

# Create a column of wells
well(1, 1:10)

# Create a matrix of wells
well(1:10, 1:10, matrix = TRUE)

# Create wells where each contains a letter
well(1:10, 1, content = letters[1:10])

# Create wells where every well contains the first 10 letters
well(1:10, 1, content = list(letters[1:10]))

# Create a matrix of wells where each well contains a number
well(1:10, 1:10, content = 1:100, matrix = TRUE)

# 1 11 ...
# 2 12 ...
# 3 13 ...
# 4 14 ...
# : :  ...

# Fill matrix by row:
well(1:10, 1:10, content = 1:100, matrix = TRUE, byrow = TRUE)

# 1  2  ...
# 11 12 ...
# 21 22 ...
# 31 32 ...
# :  :  ...

well(1:10, rep(c(1, 2), each = 5), content = 1:10)
# 1 3 5 7 9
# 2 4 6 8 10

well(1:10, rep(c(1, 2), each = 5), content = 1:10, byrow = TRUE)
# 1 2 3 4 5
# 6 7 8 9 10

These should probably error:

# Mismatch, non-multiple content
well(1:10, 1, content = letters[1:9])

# Likewise:
well(1:10, 1:10, content = 1:99, matrix = TRUE)
well(1:10, 1:10, content = 1:9)

kronecker allows for some pattern (X) to be 'dilated' to dimensions given by Y. It also allows for 'tiling' using a vector of 1s (column or row - just need to use t)