jph@suns.UMD.EDU (J. Patrick Harrington) (09/11/90)
Thanks for the replies - I'm glad to see there are a few Dyalog
users on the net, and that I was not overlooking something much
simpler. I first tried the "each" operator, which did work for
positive numbers without exponents. For the more general case,
I've only been able to do the conversion line by line. The
following function does what I need at present. Let the unix file
'NAME' have N rows with J numbers in each (assuming no more than
10000 rows); then "UR" puts the numbers into the array Z of shape
N by J:
Z <- UR NAME;R;N;J;I
[1] 'unixfiles'[]SH'unixfiles' 'open' 'tread' 'close'
[2] 'xutils'[]SH'xutils' 'ss'
[3] R <- 2 tread (N <- open NAME) 10000 & close N
[4] J <- 1 cel dc shape execute ss(ss(ss(dc R[1])'-' '@')'e' 'E')'+' ''
[5] N <- dc shape R & Z <- N J reshape 0 & I <- 1
[6] ONE: Z[I;] <- execute ss(ss(ss(I pick R)'-' '@')'e' 'E')'+' ''
[7] ->(N ge I <- I+1)/ONE
where dc=disclose, @=high minus, ge=greater or equal,
&=diamond, and cel=ceiling. Line [5] looks at the first row to
find the number of columns in the file.
Still, it would be nice to have something "smarter", that
would fill short rows with zeros, discard non-numeric comments,
etc.
J.P. Harrington
jph@astro.umd.edu
news@usc.edu (09/12/90)
In article <7273@umd5.umd.edu> jph@suns.UMD.EDU (J. Patrick Harrington) writes: Thanks for the replies - I'm glad to see there are a few Dyalog users on the net, and that I was not overlooking something much simpler. ... Z <- UR NAME;R;N;J;I [1] 'unixfiles'[]SH'unixfiles' 'open' 'tread' 'close' [2] 'xutils'[]SH'xutils' 'ss' [3] R <- 2 tread (N <- open NAME) 10000 & close N [4] J <- 1 cel dc shape execute ss(ss(ss(dc R[1])'-' '@')'e' 'E')'+' '' [5] N <- dc shape R & Z <- N J reshape 0 & I <- 1 [6] ONE: Z[I;] <- execute ss(ss(ss(I pick R)'-' '@')'e' 'E')'+' '' [7] ->(N ge I <- I+1)/ONE where dc=disclose, @=high minus, ge=greater or equal, &=diamond, and cel=ceiling. Line [5] looks at the first row to find the number of columns in the file. Still, it would be nice to have something "smarter", that would fill short rows with zeros, discard non-numeric comments, etc. First off, I'm not a Dyalog user, so maybe I should just shut up... However, normally, the way you 'fill short rows with zeros' is with an overtake. e.g. on line [6] ONE: Z[I;] <- J take execute .... As for discarding non-numeric comments, well.. that depends on what a comment is, but try: [5.5] B <- J reshape 0 [6] ONE: T <- ss(ss(ss(I pick R)'-' '@')'e' 'E')'+' '' [6.1] ->(~ v/ '0123456789' element_of T)/NEXT [6.2] ->(~ ^/ T element_of '@0123456789E ')/NEXT [6.3] B[I ] <- 1 [6.4] Z[I;] <- J take execute T [7] NEXT: I <- I + 1 [8] ->(N ge I)/ONE [9] Z <- B /[1] Z note that element_of is the old familiar epsilon, and that ~ v/ may be done more quickly as ~ 1 element_of, and that ~ ^/ may be done more quickly as 0 element_of 'nuf said
adam@ste.dyn.bae.co.uk (Adam Curtin) (09/12/90)
In article <7273@umd5.umd.edu> jph@suns.UMD.EDU (J. Patrick Harrington) writes: >Thanks for the replies - I'm glad to see there are a few Dyalog >users on the net, and that I was not overlooking something much >simpler. I'm glad, too - I worked for Dyadic until January, and unixfiles and xutils are about 90% my code (the rest being due to Martin Tann). Your problem may have a different solution with Dyalog APL version 6, which replaces unixfiles with nfiles (all mine!), which has very sophisticated input and output translation - using the apltrans and aplkeys tables. It's easy to define a pair of translation tables which will do the translations from ASCII numeric text to APL numeric text for input, and the reverse translation on output. Then all you need to do is ravel, execute and reshape! In the absence of nfiles, and not wanting to put you off using my program (!), don't neglect monadic []SH. The right argument can be any shell command line, whose output is captured (if the result is assigned) to a vector of vectors, one element per line of output. This opens up the use of tr, sed, awk, pipelines of uniq, sort, pr, ... the sky's the limit. Adam PS As a thinly-disguised sales pitch, Dyalog APL version 6 also has 'jade' (just another development environment) with windowy editing, execution trace and variable watching (jade due to yours truly and John Scholes). Version 6 is the first version of Dyalog APL to be available for DOS (on 386 or better). PPS Dyadic are now at Riverside View, Basing Road, Old Basing, Basingstoke, Hampshire RG24 0AL, UK. Tel: +44 256 811125; Fax: +44 256 811130. -- A. D. Curtin . Tel : +44 438 753430 British Aerospace (Dynamics) Ltd. . Email: adam@ste.dyn.bae.co.uk PB 230, PO Box 19, Six Hills Way, . <This disclaimer conforms to RFC 1188> Stevenage, SG1 2DA, UK. . "My other car is an FJ1200"
rjfrey@kepler.com (Robert J Frey) (09/13/90)
In article <1990Sep12.094140.107@ste.dyn.bae.co.uk> adam@ste.dyn.bae.co.uk (Adam Curtin) writes: > >I worked for Dyadic until January, and unixfiles and xutils >are about 90% my code (the rest being due to Martin Tann). > A great piece of work! > >In the absence of nfiles, and not wanting to put you off using my program (!), >don't neglect monadic []SH. The right argument can be any shell command line, >whose output is captured (if the result is assigned) to a vector of vectors, >one element per line of output. This opens up the use of tr, sed, awk, >pipelines of uniq, sort, pr, ... the sky's the limit. > This is a great environment to work in. We often prototype in APL, but are easily able to integrate normal UNIX tools via monadic quadSH. We then identify code we want to implement as AP's and prototype them as shell scripts or "quick-and-dirty" programs, integrated with monadic quadSH and/or ASCII files written and read with unixfiles and xutils. Finally, the required code is implemented as standard AP's and we use dyadic quadSH so they appear in the APL workspace itself as functions. Obviously, in the early stages of this process we make extensive use of monadic quadSH and the unixfiles and xutils AP's. This iterative refinement of imple- mentation details allows us to focus on basic design issues early on and helps us get a working system up and running very quickly and continue to improve it steadily as time goes on. On the research side, because so much code is experimental or run-once-then- throw-away, having a method for linking existing tools into APL with a trivial amount of work is important. Also, many of our researchers are hired for their mathematical, not programming skills, and quadSH, unixfiles and xutils are easier to use than AP's (though not as efficient). > >Dyadic are now at Riverside View, Basing Road, Old Basing, Basingstoke, >Hampshire RG24 0AL, UK. >Tel: +44 256 811125; Fax: +44 256 811130. > -- Dr. Robert J Frey, Kepler Financial Management, Ltd. rjfrey@kepler.com *or* frey@chaos.sunysb.edu voice: (516) 689-6300 * fax: (516) 751-8678