R, ggplot2: change fill value to discreet color instead of gradient

Question

I have the following data:

Text refers to a narrative, mi is the count of a particular feature, and shi is the count of another feature. I have them in a data frame as follows. I'm not sure how manually recreate this dataframe with vectors of the values, so I apologize in advance that this isn't as reproducible as it otherwise could be.

test <- as.data.frame(read_excel("plotTEST.xlsx"))

I want a plot in which the x axis is each text, and y shows the counts of both features. I want to approximate something like this, as shown here or here .

Using those examples I cobbled together the following:

ggplot(data=test, aes(x=text, y=mi, fill=shi)) + geom_bar(stat="identity")

And through lots of tweaking and several iterations, this is the cleanest result I have gotten:

Gradient seems to be default despite the fact that I tried following the code in the linked examples and they don't give specific details about the color selections.

It's a gradient rather than discreet colors, how can I get discreet colors for the mi counts and shi counts?

I tried this solution mentioned here but that just results in another separate key with different colors and each bar gets a different colored outline.

Answer 1

You can do it like this:

# Your data
text <- c("Swollen River", "Anaconda caught", "Chikwan speaks", "Lake explodes","Swollen River", "Anaconda caught", "Chikwan speaks", "Lake explodes")
values  <- c(5,18,45,98, 0,9,33,23)
type <- c("mi","mi","mi","mi","shi","shi","shi","shi")
df <- data.frame("text"=text, "values"=values, "type"=type)

# dataframe in long format:
#             text values type
# 1   Swollen River      5   mi
# 2 Anaconda caught     18   mi
# 3  Chikwan speaks     45   mi
# 4   Lake explodes     98   mi
# 5   Swollen River      0  shi
# 6 Anaconda caught      9  shi
# 7  Chikwan speaks     33  shi
# 8   Lake explodes     23  shi

# Plot
ggplot(data=df, aes(x=text, y=values, fill=type)) + geom_bar(stat="identity")

In other words, you do need the data preferably in long format, as described here .