🚀 go-pugleaf

tgl+@cs.cmu.edu (Tom Lane) wrote:
> I'm thinking about designing a language in which discrete time series are a
> fundamental data type.  I have vague memories of having seen similar ideas
> before, so I'm hoping someone can point me to prior work.
> The data I want to manipulate comes as series of observations taken at
> certain intervals of time (not necessarily the same interval for each
> series).
> I want to be able to refer to such a series as a single data object;
> to perform operations on such objects (eg, to add corresponding elements
> of two series to deliver a sum series), and to define functions that work
> on such objects (integration and correlation might be examples).  I want
> the functions to be able to work incrementally as new values are added to
> each data series (lazy evaluation might help here).

Try this:

AUTHOR: Wadge William W.
TITLE: Lucid, the dataflow programming language
IMPRINT: London Academic Press 1985
SERIES: Studies in data processing (Brighton College of Technology. Automatic
          Programming Information Centre) v.22
OTHER AUTHORS: Ashcroft Edward A.

-- 
--  Jack Campin    room G092, Computing Science Department, Glasgow University,
17 Lilybank Gardens, Glasgow G12 8RZ, Scotland   TEL: 041 339 8855 x6854 (work)
INTERNET: jack@dcs.glasgow.ac.uk or via nsfnet-relay.ac.uk    FAX: 041 330 4913
BANG!net: via mcsun and uknet     BITNET: via UKACRL    UUCP: jack@glasgow.uucp

In article <Bt3Jt3.3Kn.2@cs.cmu.edu> tgl+@cs.cmu.edu (Tom Lane) writes:
>I'm thinking about designing a language in which discrete time series are a
>fundamental data type.  I have vague memories of having seen similar ideas
>before, so I'm hoping someone can point me to prior work.

There are a couple of stream-oriented languages you should check out:
Lucid, which I think was first developed at SRI, and SIGNAL, developed at
INRIA, France by Benveniste, Le Guernic, and others.

>In trying to work out these ideas, I've run into some nasty problems of
>semantics, particularly when combining series having different time bases.
>How to make a reasonably efficient implementation is not obvious either.

Given this, you'll be more interested in the SIGNAL work, I think, which
deals quite a bit with different time bases and how to determine whether
systems with different time bases are consistent.



--
Joe Buck	jbuck@ohm.berkeley.edu

In article <Bt3Jt3.3Kn.2@cs.cmu.edu> tgl+@cs.cmu.edu (Tom Lane) writes:
>I'm thinking about designing a language in which discrete time series are a
>fundamental data type.  I have vague memories of having seen similar ideas
>before, so I'm hoping someone can point me to prior work.
>
>The data I want to manipulate comes as series of observations taken at
>certain intervals of time (not necessarily the same interval for each series).
>I want to be able to refer to such a series as a single data object;
>to perform operations on such objects (eg, to add corresponding elements
>of two series to deliver a sum series), and to define functions that work
>on such objects (integration and correlation might be examples).  I want
>the functions to be able to work incrementally as new values are added to
>each data series (lazy evaluation might help here).
>
>In trying to work out these ideas, I've run into some nasty problems of
>semantics, particularly when combining series having different time bases.
>How to make a reasonably efficient implementation is not obvious either.

a. One of the main reasons that Reuters bought I.P. Sharp Associates
   (aside from their blathering about "the worth of the employees",
    who they have been butchering in periodic purges), was the
   IPSA "historical data bases", which are a large collection of
   time series data on energy, economics, transport, finance, etc.

b. These time series were stored as APL arrays, in the SHARP APL File
   System. Hmm. Still are. Reuters has not been able to figure out how
   to store them in another fashion efficiently.

c. APL has proven to be the most effective tool for manipulating,
   maintaining, correcting, and analyzing, these times series. 
   The functions you describe above sound fairly simple to do in APL 
   on time series.

d. I recommend you look at either SHARP APL/PC or J. Both are
   available free, via FTP from watserv1.waterloo.edu (cd /languages/apl/...)
   or with manuals, etc., from Iverson Software Inc, 33 Major St.
   Toronto.

e. J is an ascii-based language based on the Good Ideas of APL. 
    
   The Waterloo archive also has source code for J available. The APL
   code is unfortunately only available in binary form.


Robert Bernecky      rbe@yrloc.ipsa.reuter.com  bernecky@itrchq.itrc.on.ca 
Snake Island Research Inc  (416) 368-6944   FAX: (416) 360-4694 
18 Fifth Street, Ward's Island
Toronto, Ontario M5J 2B9 
Canada

The other source of ideas for treating discrete time
series as objects is the work done at MIT by Gary Kopec
and several other people. His thesis is:

"The Representation of Discrete-Time Signals and
Systems in Programs" May 1980. 

There has been a lot of follow up work since by
Brad Meyers and several other people. We have been 
using this work extensively in a commercial product
for several years, and have found it a powerful way
to treat data.

John McDowall
ilsjohn@smartstar.com