[R-sig-DB] RSQLite problems / questions

[Dirk Eddelbuettel]

On Fri, Jul 07, 2006 at 08:29:08AM +0100, Prof Brian Ripley wrote:
> On Fri, 7 Jul 2006, Prof Brian Ripley wrote:
> 
> >On Thu, 6 Jul 2006, Dirk Eddelbuettel wrote:
[..]
> >>Same 46 limit of 46 connections I hit earlier.
> >>
> >>This seems to be a bug in SQLite or DBI.
> >
> >The message is that R is hitting its limit (50, including 3 terminal
> >connections), so the problem is in DBI.
> 
> Actually, in package RMySQL, which has in mysqlImportFile()
> 
>   if(missing(header) || missing(row.names)){
>     f <- file(fn, open="r")
>     if(skip>0)
>       readLines(f, n=skip)
>     flds <- count.fields(textConnection(readLines(f, n=2)), sep)
>     close(f)
>     nf <- length(unique(flds))
>   }
> 
> and that never closes the connection it created.  This needs to be 
> something like
> 
>     con <- textConnection(readLines(f, n=2))
>     flds <- count.fields(con, sep)
>     close(con)
> 
> or perhaps better, use an on.exit action to close the connection.

Spot on -- thanks for finding this so quickly.  RSQLite has an identical
segment in the sibbling function sqliteImportfile(), and I applied the patch
(with substitution of txtcon for con as a variable con is already in use). in

So I suppose this takes care of one of my three questions, and would
require new uploads of RMySQL, RSQLite, ... to CRAN once David resurfaces.

Thanks, Dirk

-- 
Hell, there are no rules here - we're trying to accomplish something. 
                                                  -- Thomas A. Edison