[R-pkg-devel] socketConnection, delay when reading from

Gabor Grothendieck ggrothend|eck @end|ng |rom gm@||@com
Sat Nov 27 16:45:31 CET 2021


Whether the length is variable or not isn't relevant. The point is
whether the message is prefaced by a length or command from which the
length can be derived.  Maybe it is not and you will have to rely on
inefficient methods but in many cases protocols are designed to avoid
such problems.

On Sat, Nov 27, 2021 at 9:40 AM Ben Engbers <Ben.Engbers using be-logical.nl> wrote:
>
> No, according to the specification the minimal number of bytes that is
> returned is 2. There is no maximum. (When querying a database you never
> know on forehand how many records match the query so by definition you
> can't calculate the size of the message).
>
> In some C, C++ or Java-code I found on internet, it was possible to
> change the timeout settings so that there would be no delay. Of course
> this would have as consequence that in your code you have to deal with
> the possibility that the message has not been completely returned.
>
> In R you can set the timeout to 0 but that results in errors (at least
> on Windows)
>
> Op 27-11-2021 om 14:57 schreef Gabor Grothendieck:
> > Does the message start with a length or a command whose argument length is known
> > depending on the particular command?
> > If so first read the length or command and from that the length of the
> > remainder of
> > the message can be determined.
> >
> > On Sat, Nov 27, 2021 at 4:09 AM Ben Engbers <Ben.Engbers using be-logical.nl> wrote:
> >>
> >>
> >> Hi,
> >>
> >> I have been working on a R-client for the BaseX XML-database and version
> >> 0.9.2 is nearly finished (submitting version 0.9.0 was rejected by CRAN).
> >> Version 0.3 of RBaseX can be found here
> >> (https://cran.microsoft.com/web/packages/RBaseX/index.html).
> >>
> >> The client-server protocol specifies that the communication between the
> >> client and the database is based on a socket. The code (below) shows how
> >> I create that socket.
> >>
> >> Writing to the socket works perfect. Reading from the sockets (see
> >> second codeblock) also produces correct results. The problem however is
> >> that the timeout, as specified when initializing the socket, causes a 1
> >> second delay for every read-operation.
> >> I have experimented a lot with different settings and have been
> >> searching a lot on internet, but I can't find any method to get rid of
> >> that delay. (In C or C++ that should be easier but I have never before
> >> had any need to use those languages).
> >> The very first version of my client used a block-size of 1 when reading.
> >> That gave acceptable response times for small query-results but reading
> >> large responses from the database took very long time.
> >>
> >> Do you have any suggestions on how to cope with this problem?
> >>
> >> Ben Engbers
> >>
> >> -----------------------------
> >>       CreateSocket = function(host, port = 1984L, username, password) {
> >>         tryCatch(
> >>           {conn <- private$conn <- socketConnection(
> >>             host = "localhost", port,
> >>             open = "w+b", server = FALSE, blocking = TRUE, encoding =
> >> "UTF-8", timeout = 1)
> >>           }, error = function(e) {
> >>             stop("Cannot open the connection")
> >>           }
> >>         )
> >>
> >> -----------------------------
> >>
> >> readBin_ <- function(conn) {
> >>     chars_read <- raw(0)
> >>     rd <- readBin(conn, what = "raw", 1024)
> >>     while(length(rd) == 1024) {
> >>       chars_read <- c(chars_read, rd)
> >>       rd <- readBin(conn, "raw", 1024)
> >>       }
> >>     if (length(rd) > 0) chars_read <- c(chars_read, rd)
> >>     return(chars_read)
> >> }
> >>
> >> ______________________________________________
> >> R-package-devel using r-project.org mailing list
> >> https://stat.ethz.ch/mailman/listinfo/r-package-devel
> >
> >
> >
>


-- 
Statistics & Software Consulting
GKX Group, GKX Associates Inc.
tel: 1-877-GKX-GROUP
email: ggrothendieck at gmail.com



More information about the R-package-devel mailing list