R devel - Nov 2009 - textConnection performance quadratic (PR#14053)

Full_Name: William E. Hopkins
Version: 2.9.0
OS: Windows XP
Submission from: (NULL) (209.244.4.106)


textConnection() has quadratic performance.

A function I wrote was taking outrageous amount of time to execute on a large
character vector (small test set was used for functional development). I created
a test harness to execute the function and gather stats (system.time) for
various dataset sizes (datasets generated by sample() of very large set). If I
used textConnection() to provide input to read.csv(), the performance was
quadratic with dataset size. However, if I had the function write the character
vector to a temp file then read the data back in via read.csv, the performance
was linear.

The reason for using a textConnection() was that the character vector was within
a data frame read in via read.csv. The character vector (URLs) needed to be
parsed into separate vectors, but no mechanism exists to do that directly (that
I know of). So, I used sub() to extract the proper pieces and put commas in
between so that I can use read.csv() to read the comma-separate strings directly
into vectors.

strsplit can split by separators and strapply in the gsubfn package
can split by content.

On Mon, Nov 9, 2009 at 6:40 PM,  <bill.hopkins at level3.com>
wrote:
> Full_Name: William E. Hopkins
> Version: 2.9.0
> OS: Windows XP
> Submission from: (NULL) (209.244.4.106)
>
>
> textConnection() has quadratic performance.
>
> A function I wrote was taking outrageous amount of time to execute on a
large
> character vector (small test set was used for functional development). I
created
> a test harness to execute the function and gather stats (system.time) for
> various dataset sizes (datasets generated by sample() of very large set).
If I
> used textConnection() to provide input to read.csv(), the performance was
> quadratic with dataset size. However, if I had the function write the
character
> vector to a temp file then read the data back in via read.csv, the
performance
> was linear.
>
> The reason for using a textConnection() was that the character vector was
within
> a data frame read in via read.csv. The character vector (URLs) needed to be
> parsed into separate vectors, but no mechanism exists to do that directly
(that
> I know of). So, I used sub() to extract the proper pieces and put commas in
> between so that I can use read.csv() to read the comma-separate strings
directly
> into vectors.
>
> ______________________________________________
> R-devel at r-project.org mailing list
> https://stat.ethz.ch/mailman/listinfo/r-devel
>