[R] Setting up infile for R CMD BATCH
ilai
keren at math.montana.edu
Wed Feb 8 23:43:23 CET 2012
I'm going to declare this SOLVED. Yes, if you don't want a separate
script for batch, you will need to modify the original script so it
either readline or skips it. Here is an example:
# Save in file myTest.R
####################
# Add this local function to the beginning of your original "program"
and replace all the readline prompts with myreadline.
# The new function takes on two arguments:
# "what": the original object (in your example it was "type"<-...)
# "prompt": The same string from readline
# All it is doing is searching for "Answers.R", sourcing if available
or prompting if not.
myreadline <- function(what,prompt){
ans <- length(grep('Answers.R',list.files(),fixed=T))>0 # add a
warning for multiple files
if(ans) source('Answers.R')
else{
ret <- as.integer(readline(prompt))
assign(what,ret,1)
}
}
# here is an interactive program to square only negative values
print(x <- rnorm(1))
myreadline('type','x>0 ? 1:T,2:F \n')
print(x^type)
### END myTest.R ####
Running myTest interactivly:
> source('myTest.R')
[1] -0.3712215
x>0 ? 1:T,2:F
2
[1] 0.1378054 # -.37^2
> source('myTest.R')
[1] 0.3160747
x>0 ? 1:T,2:F
1
[1] 0.3160747 # .316^1
# Create a list of answers
> dump('type',file='Answers.R')
# run again this time with answers available
> source('myTest.R')
[1] -1.088665 # skips prompt
[1] -1.088665 # -1.088^1 (type in Answer.R ==1)
# Now you can also run as batch
$ R CMD BATCH myTest.R out.R
$ cat out.R
# ...
> print(x <- rnorm(1))
[1] -1.248487
> myreadline('type','x>0 ? 1:T,2:F \n')
> print(x^type)
[1] -1.248487
That's it. Only thing is to keep the file names in the working
directory straight.
Enjoy
On Wed, Feb 8, 2012 at 12:39 PM, Gang Chen <gangchen6 at gmail.com> wrote:
> Sorry Elai for the confusions.
>
> Let me try to reframe my predicament. The main program "myTest.R" has
> been written in interactive mode with many readline() lines embedded.
> Suppose a user has already run the program once before in interactive
> mode with all the answers saved in a text file called answer.R. Now
> s/he does not want to go through the interactive again because it's a
> tedious process, and would like to run the program but in batch mode
> with answer.R. And that's why I tried the following which didn't pan
> out:
>
> R CMD BATCH answer.R output.Rout
>
> Of couse I could rewrite a separate program specifically for batch
> mode as you suggested previously with eval(), for example. However, is
> there an approach to keeping the original program so that the user
> could run both interactive and batch mode? That probably requires
> modifying the readline() part, but how?
>
> Thanks,
> Gang
>
>
> On Wed, Feb 8, 2012 at 2:08 PM, ilai <keren at math.montana.edu> wrote:
>> Gang,
>> Maybe someone here has a different take on things. I'm afraid I have
>> no more insights on this unless you explain exactly what you are
>> trying to achieve, or more importantly why? That may help understand
>> what the problem really is.
>>
>> Do you want to save an interactive session for future runs? then
>> ?save.image and all your "answers" are in the environment. In this
>> case consider putting an "if(!exists('type') | length(type)<1 |
>> is.na(type))" before "type<- readline(...)" in your script so type
>> wouldn't be overwritten in subsequent runs.
>>
>> If your goal is to batch evaluate multiple answer files from users
>> (why else would you ask questions with readline?), then you should
>> have enough to go on with my answer and the examples in ?eval.
>>
>> Elai
>>
>>
>> On Wed, Feb 8, 2012 at 9:04 AM, Gang Chen <gangchen6 at gmail.com> wrote:
>>> Hi Elai,
>>>
>>> Thanks a lot for the suggestions! I really appreciate it...
>>>
>>> Your suggestion of using eval() and creating those answers in a list
>>> would work, but there is no alternative to readline() with which I
>>> could read the input in batch mode? I'm asking this because I'd like
>>> to have the program work in both interactive and batch mode.
>>>
>>> Thanks again,
>>> Gang
>>>
>>>
>>> On Wed, Feb 8, 2012 at 12:50 AM, ilai <keren at math.montana.edu> wrote:
>>>> Ahh,
>>>> I think I'm getting it now. Well, readlines() is not going to work for
>>>> you. The help file ?readline clearly states "In non-interactive use
>>>> the result is as if the response was RETURN and the value is ‘""’."
>>>> The implication is you cannot use it to "insert" different answers as
>>>> if you were really there.
>>>> How about using eval() instead? You will need to make the answers a
>>>> named list (or just assigned objects).
>>>>
>>>> test <- expression({
>>>> if(a>2) print('+')
>>>> else print('I got more')
>>>> b <- b+3 # reassign b in the environment
>>>> print(b)
>>>> print(c)
>>>> d^2
>>>> })
>>>> dump('test',file='myTest.R') ; rm(test)
>>>>
>>>> # make the answers.R file:
>>>>
>>>> a=5 ; b=2 ; c=2 ; d=3
>>>> source("myTest.R")
>>>> eval(test)
>>>>
>>>> # Now, from the terminal R CMD BATCH answers.R out.R
>>>> # And here is my $ cat out.R
>>>> ... flushed
>>>>> a=5 ; b=2 ; c=2 ; d=3
>>>>> source("myTest.R")
>>>>> eval(test)
>>>> [1] "+"
>>>> [1] 5
>>>> [1] 2
>>>> [1] 9
>>>>>
>>>>> proc.time()
>>>> user system elapsed
>>>> 0.640 0.048 0.720
>>>>
>>>> Would this work?
>>>> Elai
>>>>
>>>>
>>>>
>>>>
>>>> On Tue, Feb 7, 2012 at 4:05 PM, Gang Chen <gangchen6 at gmail.com> wrote:
>>>>> Suppose I create an R program called myTest.R with only one line like
>>>>> the following:
>>>>>
>>>>> type <- as.integer(readline("input type (1: type1; 2: type2)? "))
>>>>>
>>>>> Then I'd like to run myTest.R in batch mode by constructing an input
>>>>> file called answers.R with the following:
>>>>>
>>>>> source("myTest.R")
>>>>> 1
>>>>>
>>>>> When I ran the following at the terminal:
>>>>>
>>>>> R CMD BATCH answer.R output.Rout
>>>>>
>>>>> it failed to pick up the answer '1' from the 2nd line in answers.R as
>>>>> shown inside output.Rout:
>>>>>
>>>>>> source("myTest.R")
>>>>> input type (0: quit; 1: type1; 2: type2)?
>>>>>> 1
>>>>> [1] 1
>>>>>
>>>>> What am I missing here?
>>>>>
>>>>> Thanks in advance,
>>>>> Gang
More information about the R-help
mailing list