CODEWITHSUNDEEP
rdata.tableplyrsqldf
beginneR
beginneR

Reputation: 3291

select-from-where procedure using R

I have two dataframes, here's the first:

df <- data.frame(p=letters[1:3],y1=c(2,4,3))
df
  p y1
1 a 2
2 b 4
3 c 3

and the second one:

df2 <- data.frame(p=rep(letters[1:3],c(3,2,4)),y2=c(3,1,1,4,3,4,3,3,1),d=rep(1,length=9))
df2
  p y2 d
1 a 3 1
2 a 1 1
3 a 1 1
4 b 4 1
5 b 3 1
6 c 4 1
7 c 3 1
8 c 3 1
9 c 1 1

What I want to do is get those lines in df2, where for each value of p(a,b,c etc.) where d=1 (which in this case are all rows), y2 is bigger than y1 grouped by p in df.

Because this explanation probably doesn't make sense, the two lines that need to be kicked: line 1 in df2, because for a, y2=3 is greater than y1=2 in df, and line 6, because for c in df2, y has value 4, but the value for c in df is 3.

Since I'm working with data.tables, a "data.table-solution" would be nice, maybe something like:

setkey(df2,d)
df2[1,y>??,by="p"]

Upvotes: 1

Views: 130

Answers (3)

agstudy
agstudy

Reputation: 121568

You should use merge before subsetting.

Using data.table:

library(data.table)
merge(data.table(df1,key='p'),
      data.table(df2,key='p'))[d==1 & y2 > y1]
   p y1 y2 d
1: a  2  3 1
2: c  3  4 1

Using base merge:

subset(merge(df1,df2), d==1 & y2 > y1)
  p y1 y2 d
1 a  2  3 1
6 c  3  4 1

EDIT

For the data.table solution , here is better to use a join Y[X] , looking up Y's rows using X's key.(LEFT OUTER JOIN)

DF2 <- data.table(df2,key='p')
DF1 <- data.table(df1,key='p')
DF2[DF1][d==1 & y2 > y1]

   p y2 d y1
1: a  3 1  2
2: c  4 1  3

Upvotes: 6

r.bot
r.bot

Reputation: 5424

Similar to rmk but using plyr:

library(plyr)
dfa <- data.frame(p=letters[1:3],y1=c(2,4,3))
dfa

dfb <- data.frame(p=rep(letters[1:3],c(3,2,4)),y2=c(3,1,1,4,3,4,3,3,1),d=rep(1,length=9))
dfb

dfb <- join(dfa, dfb, by = "p", type = "left", match = "all")
dfb

dfb$z <- ifelse(dfb$y2>dfb$y1, 1, 0)
dfb[dfb$z==1, ]

Upvotes: 1

harkmug
harkmug

Reputation: 2785

Try:

df3 <- merge(df,df2,by=1)
> df3
  p y1 y2 d
1 a  2  3 1
2 a  2  1 1
3 a  2  1 1
4 b  4  4 1
5 b  4  3 1
6 c  3  4 1
7 c  3  3 1
8 c  3  3 1
9 c  3  1 1


> df3[df3$y2>df3$y1 & df3$d==1,]
  p y1 y2 d
1 a  2  3 1
6 c  3  4 1

Upvotes: 0

Related Questions