>>>>> "Szymon" == Szymon <············@o2.pl> writes:
Szymon> How often do _you_ use SERIES?
I use it. Sometimes I use plain loop or do. Sometimes I use series.
The nice thing about series is that I can define a function that takes
a series inputs and produces series, and when I use it in some other
function, there is no function call---it's all nicely inlined.
Ray
Raymond Toy <···········@ericsson.com> writes:
> The nice thing about series is that I can define a function that
> takes a series inputs and produces series, and when I use it in some
> other function, there is no function call---it's all nicely inlined.
I find that to be the most tantalizing and maddening aspect of
SERIES. There are so many constraints that block this
optimization. It's quite difficult to avoid wandering outside of the
small world of "online ports."
Can you share some example functions that take and return series?
--
Steven E. Harris
"Steven E. Harris" <···@panix.com> writes:
> I find that to be the most tantalizing and maddening aspect of
> SERIES. There are so many constraints that block this
> optimization. It's quite difficult to avoid wandering outside of the
> small world of "online ports."
Usually I hit this problem when I'm trying to arbitrarily traverse a
series. In this case, you are using series as streams of data, not as
filters composited together.
I did the ITA programming puzzles using SERIES and I'm considering using
it for a generic lisp reader that will be used by DefEditor (so that it
reads lisp microsyntax itself as an AST -- keeping comments and other
read-macros in the code and not interning symbols when they are read).
This reader can also be used to restrict the legal structure of lisp
forms that can be read so that, for example, one can specify that a list
of a number and a symbol is to be read and any other syntax is to be
treated as an error.
--
Rahul Jain
·····@nyct.net
Professional Software Developer, Amateur Quantum Mechanicist
Rahul Jain <·····@nyct.net> writes:
> Usually I hit this problem when I'm trying to arbitrarily traverse a
> series. In this case, you are using series as streams of data, not
> as filters composited together.
Can you elaborate on the difference between streams and composited
filters? Or, rather, could you explain the latter? I know what streams
are, and, yes, that is usually how I think of series, but your comment
suggests that I may have the wrong concept in mind.
--
Steven E. Harris
[Sorry for the late reply. I've been on vacation.]
"Steven E. Harris" <···@panix.com> writes:
> Rahul Jain <·····@nyct.net> writes:
>
>> Usually I hit this problem when I'm trying to arbitrarily traverse a
>> series. In this case, you are using series as streams of data, not
>> as filters composited together.
>
> Can you elaborate on the difference between streams and composited
> filters? Or, rather, could you explain the latter? I know what streams
> are, and, yes, that is usually how I think of series, but your comment
> suggests that I may have the wrong concept in mind.
Streams of data are simply arbitrary objects that give you a datum when
asked for it. The special and optimizable case is that of a compositing
of filters, where one takes the input stream of data, transforms each
datum, possibly removing some of the data in the process, and then
combining all resulting data into some result.
--
Rahul Jain
·····@nyct.net
Professional Software Developer, Amateur Quantum Mechanicist
Szymon wrote:
>
> How often do _you_ use SERIES?
>
It's on my to-do list, but lack of asdf-installability is preventing me from
trying it out on a whim.
Svein Ove Aas <·········@aas.no> writes:
> Szymon wrote:
>
> >
> > How often do _you_ use SERIES?
> >
> It's on my to-do list, but lack of asdf-installability is preventing
> me from trying it out on a whim.
Postings like this makes me feel like an old codger. Soon I'll be
talking about keying programs in through a front panel of toggle
switches and editing paper tape with my teeth. :-)
--
Fred Gilham ······@csl.sri.com
"Due to inclement weather and massive amounts of ice everywhere,
tonight's Healthy Environment Forum on Global Warming with Dr. Patz
has been postponed", wrote Sarah Doll of the Oregon Environmental
Council in an email sent to the press, "Sorry for any inconvenience
and hope you are staying warm." --- John Daly
Svein Ove Aas <·········@aas.no> writes:
> It's on my to-do list, but lack of asdf-installability is preventing me from
> trying it out on a whim.
Getting the Debian package� should be close enough. Just unpack it and
make the symlink from asdf:*central-registry*.
Footnotes:
� http://packages.debian.org/unstable/devel/cl-series
http://ftp.debian.org/debian/pool/main/c/cl-series/cl-series_2.2.7-1.tar.gz
--
Steven E. Harris
Steven E. Harris wrote:
> Svein Ove Aas <·········@aas.no> writes:
>
>> It's on my to-do list, but lack of asdf-installability is preventing me
>> from trying it out on a whim.
>
> Getting the Debian package� should be close enough. Just unpack it and
> make the symlink from asdf:*central-registry*.
>
For that matter, I could download the upstream package and install it
myself, at even less cost.
The fact that I prefer to chat about it here instead should tell you
something about my motivation levels.
Svein Ove Aas <·········@aas.no> writes:
> For that matter, I could download the upstream package and install
> it myself, at even less cost.
My point in mentioning the Debian package was that, if I recall
correctly, the upstream package does not include an ASDF file. That's
not quite as nice as it being ASDF-INSTALLable, but being at least
ASDF-friendly makes it closer to being convenient.
--
Steven E. Harris
>>>>> "Svein" == Svein Ove Aas <·········@aas.no> writes:
Svein> Szymon wrote:
>>
>> How often do _you_ use SERIES?
>>
Svein> It's on my to-do list, but lack of asdf-installability is preventing me from
Svein> trying it out on a whim.
That can be easily remedied. But it's only 2 files: one to define
the series package, and one for the series code itself.
Ray
>>>>> "Mark" == Mark McConnell <···············@yahoo.com> writes:
Mark> Raymond Toy <···········@ericsson.com> wrote in message news:<···············@edgedsp4.rtp.ericsson.se>...
>> >>>>> "Svein" == Svein Ove Aas <·········@aas.no> writes:
>>
Svein> Szymon wrote:
>> >>
>> >> How often do _you_ use SERIES?
>> >>
Svein> It's on my to-do list, but lack of asdf-installability is preventing me from
Svein> trying it out on a whim.
>>
>> That can be easily remedied. But it's only 2 files: one to define
>> the series package, and one for the series code itself.
>>
>> Ray
Mark> How many different implementations of SERIES do people tend to use?
Mark> If there's more than one, are there significant incompatibilities, say
Mark> compared to the version in the appendix to CLtL2?
I believe there is only one, and that is the one written by Richard
Waters, which is the one in sourceforge.
Ray
Svein Ove Aas <·········@aas.no> wrote in message news:<············@services.kq.no>...
> Szymon wrote:
>
> >
> > How often do _you_ use SERIES?
> >
> It's on my to-do list, but lack of asdf-installability is preventing me from
> trying it out on a whim.
You've motivated me; SERIES 2.2.7 (the latest official release) is now
asdf-installable. Enjoy.
The asdf-packaging project STRIKES AGAIN!
-Peter
I've tried it out twice...
But haven't yet managed to see the light, in how it can improve my programs
over using normal looping constructs.
I suspect if I could see a medium sized application written with series
instead of loop, it would help a lot.
According to the series papers these exist, but I haven't ever managed to
get hold of any source code.
Both times I've reverted back to normal loops, as the pain hasn't been worth
it.
Rene.
"Szymon" <············@o2.pl> wrote in message
···················@eva.rplacd.net...
>
> How often do _you_ use SERIES?
>
> Regards, Szymon.
"Rene de Visser" <··············@hotmail.de> wrote:
> I suspect if I could see a medium sized application written with series
> instead of loop, it would help a lot.
I wrote a program once that calculated square roots using Heron's
method. I originally used something with labels and some easy
recursion. I don't know how I would have written it with loop, but I'm
sure it's possible.
Later, I went back to that code and wrote it with SERIES. You can
think of Heron's method of square root finding as an infinite series
of progressively better approximations. I translated that concept
directly into SERIES, with a simple generator function. For any
practical purpose you have to know when to stop, and SERIES
conventiently let me specify a function for this. I stopped when the
approximation was within epsilon (0.0000001 or some such small number)
of the correct square root. This is the sort of thing that seems to
fit the SERIES model well, and SERIES took all this, rearranged
everything cleverly for efficiency, and made it easier to debug. Your
milage may (and probably will) vary depending on the type of problem,
but for certain types of problems, SERIES is very nice.
-Peter
Szymon wrote:
>
> Many thanks for the replies.
>
> It seems to be that almost nobody use series...
>
For what it's worth, the asdf-ized version mentioned otherthread doesn't
compile under sbcl/slime, while loading it by hand *does* but throws
signals from redefining constants.
I'm looking into it; I'll send whoever wrote the asdf version a patch
if/when I figure out why.
Svein Ove Aas <·········@aas.no> writes:
> Szymon wrote:
>
> >
> > Many thanks for the replies.
> >
> > It seems to be that almost nobody use series...
> >
> For what it's worth, the asdf-ized version mentioned otherthread doesn't
> compile under sbcl/slime, while loading it by hand *does* but throws
> signals from redefining constants.
>
> I'm looking into it; I'll send whoever wrote the asdf version a patch
> if/when I figure out why.
It's probably about (name conflicts) EXTENSIONS:COLLECT macro and similar
stuff.
Afair SERIES documentation is not "complete" -- "internals" (like SPREAD)
are useful -- use the source to inform yourself (btw, cltl2 is good
introduction).
Regards, Szymon.
Svein Ove Aas <·········@aas.no> writes:
> For what it's worth, the asdf-ized version mentioned otherthread doesn't
> compile under sbcl/slime, while loading it by hand *does* but throws
> signals from redefining constants.
>
> I'm looking into it; I'll send whoever wrote the asdf version a patch
> if/when I figure out why.
My 2 cents: I wrote some patches for SBCL+Series to get Series into
Gentoo: http://bugs.gentoo.org/show_bug.cgi?id=57326
Kind regards,
Dirk Gerrits
>>>>> "Dirk" == Dirk Gerrits <····@dirkgerrits.com> writes:
Dirk> Svein Ove Aas <·········@aas.no> writes:
>> For what it's worth, the asdf-ized version mentioned otherthread doesn't
>> compile under sbcl/slime, while loading it by hand *does* but throws
>> signals from redefining constants.
>>
>> I'm looking into it; I'll send whoever wrote the asdf version a patch
>> if/when I figure out why.
Dirk> My 2 cents: I wrote some patches for SBCL+Series to get Series into
Dirk> Gentoo: http://bugs.gentoo.org/show_bug.cgi?id=57326
It would be useful to report these bugs (and fixes) to the series
maintainers on sourceforge.
Ray
>>>>> "Szymon" == Szymon <············@o2.pl> writes:
Szymon> Many thanks for the replies.
Szymon> It seems to be that almost nobody use series...
That isn't the conclusion that I would draw.
FWIW, I use SERIES for my stock market simulation code. Works great, the
code is less error-prone, it is also shorter and more expressive. And
quite efficient.
I did find the learning curve for SERIES to be rather steep.
As a side note, for those who have read Graham's papers, I find
CL+SERIES to be slightly higher on the language continuum than CL
itself. It amuses me that many of the arguments against series are of
the "Blub paradox" kind: "you don't need it because there's loop and
do".
--J.