The APL Orchard

dzaima

23:00

@Adám but it would fail if it's missing wherever it was executed

and better error sooner than later

nathan rogers

whatver, into the load function it goes

dzaima

(well i guess in the case of namespaces it wouldn't error sooner because Reasons™)

Adám

Still. Feel free to do so, but I'm telling you it is a bad idea. Just use a tradfn that loads the data upon first use, and/or and initialisation function that loads all when called.

rak1507

if the file doesn't exist, trying to read it at a later date is just delaying the inevitable

Adám

Maybe I can go as far as to say that namespace scripts are terribly hacky. Don't use them unless you have to. Or feel like it ;-)

ngn

23:02

just install linux and use #!

Adám

Or wait for 18.1 and use #! anywhere.

phantomics

Hi, a question for those familiar with the [↑ mix] algorithm: what function is used to reorder the dimensional bases for transforming row-major indices of the outer and inner arrays? That may not be clear. so this paste better illustrates the question: dpaste.com/627MXG8QT

Adám

Another reason not to do this load-on-fix stuff, is that debugging will be close to impossible.

@phantomics Oh, you mean mix with axis? ↑[ax]

phantomics

Yes, mix with axis

In the paste, the 0 1 2 numbers on the left show the axis that mix is operating on

dzaima

@Adám I'd say that's very much a fixable thing though

phantomics

23:05

Like ↑[0.5], ↑[1.5], ↑[2.5]

Adám

Dyalog Webinars: Language Features of Dyalog version 18.0 in Depth - Part 5

►

dzaima

@Adám (is .apln somehow special over other :namespace declarations?)

Adám

@dzaima Yes. Link reacts to .apln files.

dzaima

@Adám but iirc it also reacted to a .dyalog declaring a namespace?

Adám

@dzaima Indeed. And to .aplf and .aplo etc. too.

dzaima

23:18

do they all behave the same?

Adám

with default settings yes (except .apla)

dzaima

ok so code execution in :namespace is as it always has been (i.e. it's a poor man's #!/script)

Adám

Kind of.

@dzaima Because Link doesn't actually need to know the contents (except for arrays). It is all handed off to 2⎕FIX to figure out.

dzaima

(also today finally decided to work more on my matrix bridge, have been chatting from it :D)

(also from a custom chat client because why not)

Adám

@dzaima They were never intended to be used like #! scripts, so YMMV if you try to abuse them for that.

In fact, out CEO was rather upset when she found out :Namespace allows arbitrary code execution at fix time.

dzaima

23:23

@Adám besides the issue about multiline dfn operands, are there any other problems? (isn't a namespace what TIO uses?)

Adám

@dzaima Yes. Plenty.

dzaima

@Adám and will #! scripts inherit those, or will it be a whole another evaluation mode (to add to the list of the current ±∞)?

Adám

@dzaima Nether. #! is just allowing a text file to supply what you'd type into the session (but with multi-line support and separated input stream for ⎕/⍞).

dzaima

@Adám would anything break if :namespaces interpreted their contents like that too?

rak1507

I was hoping that #! scripts were an opportunity to remove some of the bad stuff ⎕IO, ⎕ML, ⎕DIV... but not surprising that that's not the case

Adám

23:28

@dzaima Possibly not. Interesting idea. Though classes are built on the same system, and they need special treatment.

@rak1507 Heh, you might be surprised!

rak1507

:O

dzaima

@Adám do classes allow arbitrary code execution?

Adám

yes, just like nss

ngn

hehe, the last 2 messages taken without context sound like you're talking about vulnerabilities in nss (the tls library) :)

nathan rogers

nss? @Adám

Adám

23:32

NameSpaceS

@rak1507 They are not being outright removed, but the, imo, worst aspect of those is that you never know what kind of settings your code will be run with, because the end-user computer may have envvars, reg settings, cmdline args, etc. So for #! scripting, we ignore all those.

rak1507

but you can still set them in the script?

Adám

Yes, but you only need to set those you want non-default.

rak1507

makes sense

Adám

In regular code, you have to set all those that will affect your code, so almost always ⎕IO and ⎕ML and often others too.

rak1507

still think getting rid of ⎕ML in particular might encourage people to port legacy code 'want to use this feature? update your code! insert evil laugh here'

nathan rogers

23:35

@Adám why

Adám

Because it means you can't safely load code to take a look at it. It may erase your HDD.

dzaima

@nathanrogers it's like if a C header file printed to stdout on importing or something

Adám

@rak1507 It'd just make people not port.

rak1507

@Adám there's no reason they have to use #! scripts

nathan rogers

@dzaima you've heard of JAI? That's exactly what the preprocessor does

Adám

23:37

@rak1507 Right. In fact, why would they?

nathan rogers

@dzaima the difference being that the loading and linking happens at compile time

often times you DO want to put out some kind of debug text as part of your precompiler

because what is a compiler if not just another program

I can print in python libraries on import

rak1507

@Adám they probably wouldn't, but if they ever did want to use it then having to at modernise some code might be worth it for them

nathan rogers

or load data on import

rak1507

or at least if writing any new code, write it in a more modern way

Adám

@rak1507 If we were to remove ⎕IO in #! scripts, what would it mean for code that the script would load and run?

dzaima

23:39

@nathanrogers I'm not disagreeing that there should be a way to do things at import time, but more trying to explain the actual purpose of :namespace

nathan rogers

the purpose of namespaces in programming is to create... spaces... for names

otherwise normal bodies and blocks of code

rak1507

@Adám true...

nathan rogers

but they share.... the space.... of names

dzaima

@nathanrogers right. And not evaluate code.

nathan rogers

i mean

yes and evaluate code

rak1507

23:40

how is func ← {thing} not evaluating code?

dzaima

@rak1507 well, it is. ¯\_(ツ)_/¯

nathan rogers

like are people just setting variable names to constant ints and charvectors and THATS IT?

here, I have a whole space for names, but I'm not allowed to name anything that might contain evaluation

lol

ngn

@nathanrogers you insisted on using dyalog, didn't you? get used to this kind of stuff

nathan rogers

I mean, name me a more viable APL

ngn

i can't (unless you consider j, k and the similar to be apls)

Adám

23:47

Sometimes, the conversation here sounds like people are being forced to use Dyalog APL. Really odd.

(And no, none of the people that make it sound that way are locked into jobs that have them use Dyalog APL.)

nathan rogers

Thanks for that clarification

ngn

@Adám assuming they want to use apl, yes they are kind of forced, though it's not you (dyalog) who forces them

nathan rogers

When there are no other viable alternatives

yeah it certainly does seem that way

When the thing that "those people" like has problems that make it difficult to use, yeah you'd think "those people" have a vested interest in those things being fixed

odd, that someone who wants to use a tool wants the tool to work consistently and in a variety of contexts

Adám

@ngn How so? Are they not free to use another APL impl.?

ngn

4 mins ago, by nathan rogers

I mean, name me a more viable APL

nathan rogers

23:51

⍥

Adám

@nathanrogers No, but odd that someone who wants to use a tool speaks as if they are forced to use that tool.

ngn

@Adám "assuming they want to use apl"

nathan rogers

that's your projection because it isn't the content of the text in question

Adám

"I dislike using this tool. Fix the tool for me!"

nathan rogers

more like "i'm having a problem how do I move past it"

seems like a pretty bog standard sort of question

what isn't bog standard is ignoring that it is an issue, and deflecting

rak1507

23:52

thenib.com/wp-content/uploads/2019/08/… replace society with APL

phantomics

I opted for the "write your own" solution, though that's not for most people

Adám

@nathanrogers Way more money is made using non-Dyalog APLs than Dyalog APL.

phantomics

Which ones? APL2? I heard IBM just sold it off

nathan rogers

that's fine, I'm not making money with APL, so

Adám

@phantomics And APL+

phantomics

23:55

Thought I heard that APL+ is basically orphan now after STSC went bankrupt

Adám

The people that are annoyed enough with Dyalog APL (or any thing, really) to not use it are free to leave. Funny enough, none of those vendor locked-in people with huge existing code bases, are generally the least likely to complain about aspects of the implementation.

@phantomics apl2000.com

ngn

@Adám this is the apl orchard, not the dyalog orchard, right?

Adám

@ngn Correct.

nathan rogers

probably

because they're all over 80

Adám

@ngn I didn't mean leave the Orchard. I meant leave Dyalog APL behind.

ngn

23:58

@Adám ah, ok

phantomics

Yeah, APL2000 seems to still be healthy, enough that they've told off hobbyists who've asked if they could release an evaluation edition

Do you have any idea what verticals are using it? Insurance?

Adám

@ngn Should have been clear from me mentioning "any thing".

@phantomics No, it is all very secretive.

phantomics

Interesting

ngn

@Adám s/free to leave/free to leave it/

Adám

@ngn Fair. Sorry.

ngn

23:59

no problem

Transcript for