Re: Backspacing in input

In article <C76M6H.111@promdist.pds.com> bill@promdist.pds.com (Bill Foote) writes: >In article <C6yFps.85C@informatik.uni-muenchen.de> spitz@ana.med.uni-muenchen.de writes: >> >>One other SINIX enhancement in 2.1 that was dropped in 4.0 was the >>load from/unload to PIPE. I really liked that nifty feature, since >>it allowed statements like >> >> load from pipe "rsh remotehost file | awk 'something' | sed 'anything' ">> >>and so on. Is there a chance to include that in future versions of >>INFORMIX? > >Does SINIX have named pipes? If so, they could achieve the same effect. > >In case you're not familiar with them (I just learned about them myself >from the recent thread on unattended backups :-), here's how to do what >you want in SCO. (In SCO, named pipes are created with "mkfifo", and "rsh" >is called "rcmd"). > > > <from the shell> > mkfifo /tmp/my_fifo > rcmd remotehost "file | awk 'something' | sed 'anything'" > /tmp/my_fifo & > isql > load from pipe "/tmp/my_fifo" insert into ...> <exit isql> > rm /tmp/my_fifo > One more change: reader before writer in most versions of UNIX I know. >This is not an excuse for "load from pipe ..." not being part of isql. One of the difficulties is in detecting the difference betwen a char string for a file (type p) and for a _proces_pipeline: cmd | cmd | cmd could also be: lokslikefileinpwd_BUT_iscmdfromPATH ... so that if you really want flexibility, you have to distinguish the two. I believe the choice they are making is: 1) for proces invocation use 'load from pipe' 2) for any type of file oriented input, including named or unnamed pipe standard in or open FD b/c/r type device streams pipe devices use "load from" or "load from file" May not be there entirely yet, but I expect you will see it son. Also, to use streaming files properly, the code inside OL may have to be altered to do very strict "read once sequentially"; it may in fact not do this. There are other cmds (such as tbape) which do a doublke pump on the stream at hand; it requires very careful coding (actually, a unifr\\orm front end IO subroutine with buffering, pushback, etc. handled, an\\bout 10 pages of VERY careful coding) to do these things write (errr, right)> ANd there are some exceedingly weird situations that have to be handled in regards to isolation, stable data and rollback. Not completely trivial. roy andrew crabtree uunet!mcnc!rbdc!royc >It seems like an obvious feature (and it should only take Informix a line >or two of code to implement). >-- >Bill Foote bill@pds.com >PDS Corp, Los Angeles, CA USA uunet!promdist!bill