using hash to determine whether 2 dataframes are identical (PART 01)

问题

I have created a dataset using WHO ATC/DDD Index a few months before and I want to make sure if the database online remains unchanged today, so I downloaded it again and try to use the digest package in R to do the comparison.

The two dataset (in txt format) can be downloaded here. (I am aware that you may think the files are unsafe and may have virus, but I don't know how to generate a dummy dataset to replicate the issue I have now, so I upload the dataset finally)

And I have written a little script as below:

library(digest)

ddd.old <- read.table("ddd.table.old.txt",header=TRUE,stringsAsFactors=FALSE)
ddd.new <- read.table("ddd.table.new.txt",header=TRUE,stringsAsFactors=FALSE)


ddd.old[,"ddd"] <- as.character(ddd.old[,"ddd"])
ddd.new[,"ddd"] <- as.character(ddd.new[,"ddd"])

ddd.old <- data.frame(ddd.old, hash = apply(ddd.old, 1, digest),stringsAsFactors=FALSE)
ddd.new <- data.frame(ddd.new, hash = apply(ddd.new, 1, digest),stringsAsFactors=FALSE)

ddd.old <- ddd.old[order(ddd.old[,"hash"]),]
ddd.new <- ddd.new[order(ddd.new[,"hash"]),]

And something really interesting happens when I do the checking:

> table(ddd.old[,"hash"]%in%ddd.new[,"hash"]) #line01

TRUE 
 506 
> table(ddd.new[,"hash"]%in%ddd.old[,"hash"]) #line02

TRUE 
 506 
> digest(ddd.old[,"hash"])==digest(ddd.new[,"hash"]) #line03
[1] TRUE
> digest(ddd.old)==digest(ddd.new) #line04
[1] FALSE

line01 and line02 shows that every rows in ddd.old can be found in ddd.new, and vice versa.
line03 shows that the hash column for both dataframe are the same
line04 shows that the two dataframe are different

What happen? Both dataframe with the identical rows (from line01 and line02), same order (from line03), but are different? (from line04)

Or do I have any misunderstanding about digest? Thanks.

回答1:

Read in data as before.

ddd.old <- read.table("ddd.table.old.txt",header=TRUE,stringsAsFactors=FALSE)
ddd.new <- read.table("ddd.table.new.txt",header=TRUE,stringsAsFactors=FALSE)
ddd.old[,"ddd"] <- as.character(ddd.old[,"ddd"])
ddd.new[,"ddd"] <- as.character(ddd.new[,"ddd"])

Like Marek said, start by checking for differences with all.equal.

all.equal(ddd.old, ddd.new)
[1] "Component 6: 4 string mismatches" 
[2] "Component 8: 24 string mismatches"

So we just need to look at columns 6 and 8.

different.old <- ddd.old[, c(6, 8)]   
different.new <- ddd.new[, c(6, 8)]

Hash these columns.

hash.old <- apply(different.old, 1, digest)
hash.new <- apply(different.new, 1, digest)

And find the rows where they don't match.

different_rows <- which(hash.old != hash.new)  #which is optional

Finally, combine the datasets.

cbind(different.old[different_rows, ], different.new[different_rows, ])

来源：https://stackoverflow.com/questions/7306212/using-hash-to-determine-whether-2-dataframes-are-identical-part-01

标签

hash