[RkWard-devel] Second 0.4.6 preview release, testing, translations

[Prasenjit Kapat]

Hi,

I have been doing a few testing...
This mail: comments on Import CSV.
Next mail: (hopefully) comments on X11 Export.

I have numbered my comments in decreasing order of importance (IMHO).

> ------- Testing -------
> - File->Import Format->Import CSV data
> The CSV import plugin is fairly old, but I've reworked several aspects.

[] Is there any particular reason for not providing the other options in 
read.table in "Further Options"? Like, quote, stringsAsFactor, flush. Call it 
coincidence, but the first csv file I tried did not work with the default 
read.table arguments, instead read.csv worked. The issue was with the quote 
characters. There was no  double quotes in the file (") but some names 
(French! nothing personal.) included single quotes ('). As a result, it was 
not being read properly. 

Specifically, the following works:
read.table(file="file.csv",header=TRUE,sep=",", quote = "", 
stringsAsFactors=FALSE)

but the following did not:
read.table(file="file.csv",header=TRUE,sep=",")

Of course, this is NOT an issue with RKWard, it is just a matter of providing 
the right specs, which when not done, seems to an end user, as an issue with 
RKWard; so, why not include the other arguments to read.table too?

[] After "Submit"ing why is the plugin still displayed? I got confused 
initially, as to whether or not the "Submit"tion actually worked. I guess, 
once the data has been read, the plugin should close. Again, I might be 
missing some subtle point!

[] Vector of row/column names and vector of column classes: As the 
import_csv.rkh suggests, these need to be of the form 
c("var1","var2",..,"varn"). But existing character variable also works, 
right? (Yes, it does, I just checked) That is, instead of specifying the 
vector directly in the input box, I can just specify the name of an existing 
varaible which contains my list of row/column names. In that case, may be add 
a few lines to the rkh file, saying, "e.g c("var1",..,"varn") or an existing 
variable".

[] Under the "Columns" tab, I think it is more appealing to include the input 
boxes inside the corresponding frames. 

[] In import_csv.rkh, some of the statements are missing a period in the end.

[] In import_csv.rkh, the following line 
        c ("row1", "row2", ... "rown")
should read
        c ("row1", "row2", ..., "rown")
What is the difference? A comma (.) after the ellipses. Similarly for the 
column name specs. Ha, I am just nitpicking. (This is what happens when you 
write too much of technical tex stuff :()

[] Should the "CSV" be renamed to something else, since the plugin provides 
option for "\t" or  ";" or "," or " " as  the separators? Maye be, 
File->Import Format->Read.Table ?

[] When on this, may be add another radio under "Feld Separator" as "Other" 
and an input box (enabled only for "Other") for any user defined separator?

[] Now I am shooting wild: (Not for 0.4.6 release): Is it possible to provide 
a "preview" of the import much like what the various office suites (oocalc, 
kspread etc.) do, to let the user see what is actually happening/going to 
happen once the import is completed?

[] This comment might make the plugin too complicated/confusing. What I was 
thinking is: add a dropbox under the "General" tab, which will decide a 
predetermined set of defaults for the various parameters. To be precise, the 
drop box, will include: None, CSV, CSV2, DLIM, DLIM2 which in turn will the 
set the paramters to the read.csv, read.csv2, read.dlim, read.dlim2 
respectively. 
	Ok, though this is a thought of mine, but I am very uncomfortable with this 
being implemented (self contradicotry!). What do others think?

Regards
PK