Deserialise-- does it work as expected?

General APL language issues

Re: Deserialise-- does it work as expected?

Postby Phil Last on Wed Aug 10, 2022 2:26 pm

Adam|Dyalog wrote:Any way of contacting us is perfectly fine. We can continue the discussion in this forum, but we do not want to limit feedback to forum users. We will do our best to record feedback, responses and any resulting modifications on the APL Wiki page.
My question was specifically aimed at finding a central place where suggestions could be gathered and visible to all. If I make a suggestion in one place and Joe Bloggs makes a similar but incompatible suggestion elsewhere we are neither of us the wiser and consensus is neither invited nor likely.

And reference back to Peter's question to emphasise that there are inconsistencies and contentious issues that require consensus in preference to arbitrary dictation.
User avatar
Phil Last
 
Posts: 606
Joined: Thu Jun 18, 2009 6:29 pm

Re: Deserialise-- does it work as expected?

Postby Bob Armstrong on Sun Aug 14, 2022 3:38 pm

I remember the linearization , serialization , of math notation was an ` accomplishment Ken considered significant for eliminating , among other things , ambiguity .
But , the structuring of data i/o is a separate issue from notating algorithms .

Just for comparison , in CoSy https://www.cosy.com/CoSy/ RPN I'd likely write the example at https://apl-germany.de/wp-content/uploa ... df#page=34 as

Code: Select all
  (' `( Name Age Sex )` (' ` Anton `f 56 ` M ') (' ` Berta `f 93 ` F ') (' ` Cäsar `f 11 ` M ') ') >t0>

(
 (
  Name
  Age
  Sex
 ) (
  Anton
  56.0000
  M
 ) (
  Berta
  93.0000
  F
 ) (
  Cäsar
  11.0000
  M
 ) )

  t0 #
4

  t0 ' # 'm ,/
3 3 3 3

  t0 fmttbl|
(
 Name  | Anton     | Berta     | Cäsar   
 Age   | 56.0000   | 93.0000   | 11.0000 
 Sex   | M         | F         | M       
 )

Where ` returns the following ` word ( nonblank string ) as a string , `( ... )` does it on a list , ` `f converts the following number to a 1 item float list , and (' ... ') executes the string between making a list of the results .

There's a lot that's still crude about CoSy because it is still executing directly at the x86 Forth level , and it's still only been written by me , but being open Forth , it is unmatched in its ` flexibility .

Incidentally , the left tick , which I'm tending to call qua , which returns the following ` word as a string is largely influenced by Morten's comment at http://cosy.com/CoSy/MinnowBrook2011.html which caused me to eliminate a separate ` symbol type which I had carried forward from K . I found having a separate Type complexity with no advantage .
User avatar
Bob Armstrong
 
Posts: 23
Joined: Wed Dec 23, 2009 8:41 pm
Location: 39.038681° -105.079070° 2500m

Re: Deserialise-- does it work as expected?

Postby Adam|Dyalog on Mon Sep 05, 2022 12:15 pm

Phil Last wrote:My question was specifically aimed at finding a central place where suggestions could be gathered and visible to all. If I make a suggestion in one place and Joe Bloggs makes a similar but incompatible suggestion elsewhere we are neither of us the wiser and consensus is neither invited nor likely.

We do not want to try to force the discussion to a single place – especially not one that requires people to sign up before they can post comments. It also shouldn't be required to have one's comments be publicly visible.

APL Wiki now has two articles, Array notation and Array notation design considerations. The former strives to describe the notation in pure form. The latter details how we got there, and its discussion page is open for additional feedback, in a manner similar to how it was successfully done for the APL Logo. To this page, we will also add (possibly anonymised) all substantial comments and suggestions received via any other channels, for example this forum and various chat rooms, which we will monitor for any such feedback.

We currently plan on implementing the notation as described, but would like to get as much feedback as possible from the rest of the community, in the hope that we might end up with a notation that we could share, at least the core of it.

Phil Last wrote:And reference back to Peter's question to emphasise that there are inconsistencies and contentious issues that require consensus in preference to arbitrary dictation.

For the uninitiated reader: These turned out be non-issues.
User avatar
Adam|Dyalog
 
Posts: 114
Joined: Thu Jun 25, 2015 1:13 pm

Re: Deserialise-- does it work as expected?

Postby Phil Last on Tue Sep 06, 2022 10:02 am

Adam wrote:We do not want to try to force the discussion to a single place


Why?

Adam wrote:- especially not one that requires people to sign up before they can post comments.


I'm rather stuck to think of many places where signing up is not required.

Adam wrote:It also shouldn't be required to have one's comments be publicly visible.
[...]
we will also add (possibly anonymised) all substantial comments and suggestions [to the APL Wiki]

I'm also stuck to think of a single reason why anyone should want to remain anonymous or keep one's comments from public scrutiny.

Is it possible that Dyalog - contrary to advice - is trying to avoid a public consultation?

Finally, a wiki is not the place to post suggestions and comments.
User avatar
Phil Last
 
Posts: 606
Joined: Thu Jun 18, 2009 6:29 pm

Re: Deserialise-- does it work as expected?

Postby petermsiegel on Wed Sep 07, 2022 4:49 am

I think Phil's thinking aligns with members of other technical user communities I have known (vendor and otherwise) and avoids the "you didn't check webpage/wiki/site X, where this problem/solution/plan was discussed" syndrome. Here's a quick précis on what may be out there in the ecosystem. OK if [TL;DR].

Some popular, public domain languages and major toolkits started so early with enhancement proposals (Python's term, aka PEPs) and community feedback that they still use older technology for community input/involvement they wouldn't choose today. These include email lists / well-maintained web digests (Python) or features like "bug" reports/web digests to contribute comments or solutions (Ruby). Python has tremendous inertia here, without much impetus perhaps to move to a more modern tool.

Still others (e.g. node.js) use Github Issues, which seems most flexible, with positive community member feedback. Some topics may be better suited for Github Discussions.

Early observations:
    • Most (all?) seem to require an account (but not necessarily one the group controls).
    • The best approaches are a careful balance of core (perhaps vendor) leadership and a meritocracy of active users (via intellectual contributions or sweat equity), including requirements-savvy customers.
    • This all requires investigation: left as an exercise to the reader.
Having lived the model of a vendor-sponsored (but not controlled) user group in the Mesozoic era (when the original Array Processors roamed the earth), the best ones had a user team that surveyed the community broadly, both defining and "owning" the process, working with and well-supported by the vendor.

Why would a vendor sanction this? 1) Because the benefit of strong user engagement and earned loyalty outweighed the disadvantages of some users asking for stuff they wouldn't actually buy/build/use. 2) Being an underdog helps. 3) Users are very good at sorting out chaff from wheat.
Even without such a model, a cursory review of the best solutions out there beats smart people thinking hard before fully considering the lay of the land.

Cheers
petermsiegel
 
Posts: 125
Joined: Thu Nov 11, 2010 11:04 pm

Re: Deserialise-- does it work as expected?

Postby Morten|Dyalog on Wed Sep 21, 2022 10:26 am

As I understand it, Phil and others are suggesting that we pick an online forum or mechanism, and declare that it is the one location where the discussion about array notation should be happening. Our position is that we are happy to receive feedback via any channel, that we will curate the responses, write them up in a consistent manner, and post them in a location where everyone can read them without having to sign up. I will try to explain why we want to keep things open.

Most importantly, feedback has been extremely limited, despite many years of repeated presentations and requests for input. Almost exclusively, feedback has come from Phil, whose thoughts are already deeply baked into the proposal, and to whom we are extremely grateful.

Given the importance of getting this design right and the lack of feedback to date, we want to make one more attempt, casting the widest possible net before we set the implementation in stone (i.e. C code). In addition to another call for participation in webinars (Adam is presenting on the topic tomorrow at the BAA), forums and chat rooms, it is our intention to write directly to a number of key users and implementors and ask them to carefully read the proposal and tell us what they think. I fear that some of the people whose opinions we value highly will be unwilling to - or maybe even prohibited from - simply posting their thoughts on quote social media unquote. Some have ridiculed this worry, but I am convinced that the problem is real and that the issue at hand is too important to risk having the most, ah, self-confident people dominate the discussion, as is often the case in on line discussions.

I believe that we are likely to get more thoughtful and carefully written responses via private consultation rather than rapid-fire online discussions. Unfortunately, I feel that this thread illustrates only too well the typical signal-to-noise ratio of public debate, and demonstrates clearly why a curated log of objections and adjustments or responses is necessary. Maybe this is old fashioned, but when it comes to designing things that need to last for generations, I think being a little old-fashioned is a virtue.

Please believe me when I say that we are EXTREMELY interested in feedback regarding the notation, and we harbour deep anxieties about getting this wrong and having to live with it "for eternity". We have created an Array Notation Discussion topic on our Forums. We welcome input there, but also by any widely used array language channel, as well as private communications. We pledge to write up serious objections and answers on the APL Wiki for all to see.

We agree that Wiki "talk" pages are NOT a good way to discuss this kind of thing, but a Wiki is a convenient place to store the result of the curated discussion. Anyone who feels that we are not taking a significant objection seriously is welcome to shame us in public, of course.

Please read the proposal carefully, think about how you might use it in the future, and let us know how you think it might be improved!

If Adam or I, or any other member of our team, have ignored feedback, please remind us now. If we have written anything in this thread or elsewhere in the past, that has contributed to a feeling that we are not interested in feedback, please try to forgive us!
User avatar
Morten|Dyalog
 
Posts: 432
Joined: Tue Sep 09, 2008 3:52 pm

Previous

Return to Language

Who is online

Users browsing this forum: No registered users and 1 guest