R devel - Sep 2003 - How does "subset" replace arguments? (PR#4193)

Full_Name: Axel Benz
Version: 1.7.1
OS: Windows
Submission from: (NULL) (137.251.33.43)


Hello,
I guess many people will answer me again that this is a S language feature, but
I am only a stupid computer scientist and I simply do not understand this logic,
despite of reading a lot about S:
> test
   field           tuckey
4  Kreis2          -1
5  Kreis5          -2
9  Metall          -3
17 Kreis1          -4
19 Kreis8          -5
> subset(test,field=="Metall")
  field       tuckey
9 Metall      -3
> subset(test,toString(field)=="Metall")
[1] field   tuckey
<0 rows> (or 0-length row.names)

This happens everytime I use a function with the column name ("field",
in this
case) as parameter in the  logic expression in "subset", instead of
using the
column name on top level. I have the impression that the column name is only
replaced when standing in top level position. I would call that "very lazy
evaluation" ;-) ;-)
Thank you for a friendly answer, this language is realy weird to me.

On Tuesday 16 September 2003 14:39, axel.benz@iao.fhg.de
wrote:
> Full_Name: Axel Benz
> Version: 1.7.1
> OS: Windows
> Submission from: (NULL) (137.251.33.43)
> 
> 
> Hello,
> I guess many people will answer me again that this is a S language feature,
but
> I am only a stupid computer scientist and I simply do not understand this 
logic,
> despite of reading a lot about S:
> 
> > test
>    field           tuckey
> 4  Kreis2          -1
> 5  Kreis5          -2
> 9  Metall          -3
> 17 Kreis1          -4
> 19 Kreis8          -5
> 
> > subset(test,field=="Metall")
>   field       tuckey
> 9 Metall      -3
> 
> > subset(test,toString(field)=="Metall")
> [1] field   tuckey
> <0 rows> (or 0-length row.names)
I don't see any problem here. toString(field), evaluated in the data frame 
test, should be the single string

"Kreis2, Kreis5, Metall, Kreis1, Kreis8"

So, the comparison 

toString(field)=="Metall" 

actually does 

"Kreis2, Kreis5, Metall, Kreis1, Kreis8" == "Metall"

which being false, returns FALSE, and so you finally should get 

subset(test, FALSE)

which is what you do get.

Perhaps you misunderstood what the function toString() does.

HTH,

Deepayan

P.S. Please don't use R-bugs to report what may or may not be bugs, since
all
such reports have to be processed manually. Ask on r-help or r-devel first if 
you are not sure.

On Tue, 16 Sep 2003 axel.benz@iao.fhg.de wrote:
> Full_Name: Axel Benz
> Version: 1.7.1
> OS: Windows
> Submission from: (NULL) (137.251.33.43)
>
>
> Hello, I guess many people will answer me again that this is a S
> language feature, but I am only a stupid computer scientist and I simply
> do not understand this logic, despite of reading a lot about S:
The point they are trying to make is that you should send this sort of
question to r-devel or r-help, not r-bugs.  The point of r-bugs is as a
repository for bug reports, not as a discussion list.

> > test
>    field           tuckey
> 4  Kreis2          -1
> 5  Kreis5          -2
> 9  Metall          -3
> 17 Kreis1          -4
> 19 Kreis8          -5
>
> > subset(test,field=="Metall")
>   field       tuckey
> 9 Metall      -3
>
> > subset(test,toString(field)=="Metall")
> [1] field   tuckey
> <0 rows> (or 0-length row.names)
>
> This happens everytime I use a function with the column name
("field", in this
> case) as parameter in the  logic expression in "subset", instead
of using the
> column name on top level. I have the impression that the column name is
only
> replaced when standing in top level position. I would call that "very
lazy
> evaluation" ;-) ;-)
> Thank you for a friendly answer, this language is realy weird to me.
>
Your impression is incorrect.  The problem with toString is that it
collapses a vector to a single string, so toString(field) is the string
"Kreis2, Kreis5, Metall, Kries1, Kries8".  There is no record whose
`field' is equal to that string.  Did you check to see that toString did
what you thought it did?


subset() will work as I think you expect if the output of the function is
the same length as the input.

For example, consider one of the built-in data sets

data(esoph)
> subset(esoph, toString(agegp)=="75+")
[1] agegp     alcgp     tobgp     ncases    ncontrols
<0 rows> (or 0-length row.names)

but
> subset(esoph, as.character(agegp)=="75+")
   agegp     alcgp    tobgp ncases ncontrols
78   75+ 0-39g/day 0-9g/day      1        18
79   75+ 0-39g/day    10-19      2         6
80   75+ 0-39g/day      30+      1         3
81   75+     40-79 0-9g/day      2         5
82   75+     40-79    10-19      1         3
83   75+     40-79    20-29      0         3
84   75+     40-79      30+      1         1
85   75+    80-119 0-9g/day      1         1
86   75+    80-119    10-19      1         1
87   75+      120+ 0-9g/day      2         2
88   75+      120+    10-19      1         1


or to take a really extreme version
> subset(esoph,
substr(paste(as.character(agegp),toupper(as.character(agegp))),3,6)== "+
75")
   agegp     alcgp    tobgp ncases ncontrols
78   75+ 0-39g/day 0-9g/day      1        18
79   75+ 0-39g/day    10-19      2         6
80   75+ 0-39g/day      30+      1         3
81   75+     40-79 0-9g/day      2         5
82   75+     40-79    10-19      1         3
83   75+     40-79    20-29      0         3
84   75+     40-79      30+      1         1
85   75+    80-119 0-9g/day      1         1
86   75+    80-119    10-19      1         1
87   75+      120+ 0-9g/day      2         2
88   75+      120+    10-19      1         1


	-thomas