On 30 July 2013 at 11:26, Wush Wu wrote:
| Thanks for your suggestion, Dirk.
Glad to have been of assistance. What you were trying to do (hold on to and
shield a connection object), Modules are a very fine tool.
| I'll try to rewrite Rhiredis based on your approach and provide a similar
| rredis interface. If everything is good, I'll try to make it on CRAN if it is
| possible.
I'd be happy to help.
The one thing it needs is a C-level access to serialization which would help
with the overall speed gain. Right now we're not all that much faster than
the rredis package. Below is a simple benchmark (which is part of my git
repo 'rhiredis') that attempts to be honest in the comparison to rredis.
Dirk
library(rhiredis)
library(rredis)
library(rbenchmark)
data(trees)
fit <- lm(log(Volume) ~ log(Girth) + log(Height), data=trees)
rawfit <- rawToChar(serialize(fit,NULL,ascii=TRUE))
redis <- new(Redis)
rredis::redisConnect()
hiredis <- function() redis$exec(paste("SET fits1 ",
rawToChar(serialize(fit,NULL,ascii=TRUE)),
sep=""))
rredis <- function() rredis::redisSet("fits2", fit)
res <- benchmark(hiredis(), rredis(), replications=100)[,1:4]
print(res)
all.equal(unserialize(charToRaw(redis$exec("GET fits1"))), fit)
all.equal(rredis::redisGet("fits2"), fit)
--
Dirk Eddelbuettel | e...@debian.org | http://dirk.eddelbuettel.com
_______________________________________________
Rcpp-devel mailing list
Rcpp-devel@lists.r-forge.r-project.org
https://lists.r-forge.r-project.org/cgi-bin/mailman/listinfo/rcpp-devel
