hi

sorry i had issues logging in

can npm install themselves into other packages?

without scripts.

got ignore-scripts=true. Can their "unpack path" be into other modules?

is there any way to create a pending promise and pass it reference to some code that will complete it once it's ready? In current js' way, you have to define the resolution function during creation. That is, you can't create an "unbound" promise object.

> you have to define the resolution function during creation

you do?

that is, I need the Q way: `return Q.defer().promise`

cek: a 'defer' is what you're looking for, for which you can use Promise.defer if using Bluebird, or the `p-defer` module if using ES6 Promises -- though the usual warnings apply, only do this when you *really* have no other reasonable way to express your code (eg. it's task queue code), because it's easy to get wrong and hard to follow

it's a dispatcher code, a state routing machine if you will

but I got the idea, no built-in.

that sounds like it could be a valid case for defers, yeah

there was a library that would log all require calls for debugging purposes. can't find it

_rgn: isn't that doable with just the NODE_DEBUG env var?

Hello. I recently began working on a colleague's project. I reorganized it a bit, and have broken the main project test-- just a couple parts of the test. But I am getting this error and I don't quite understand why.

" getaddrinfo ENOTFOUND undefined"

I read that it may have something to do with a malformed "http" vs "https" notation or something like that...

yaalon: that's a DNS lookup error, means that a DNS request was made for a non-existent domain/hostname

malformed URLs are one possible cause of that

Thanks a bunch! Interesting

eg. if you have a URL like this: http//example.com/bar

(note the missing : colon)

then it may parse that as the domain "http" (which doesn't exist) and a path of "//example.com/bar"

In other words, if you refactored any parts of the code that builds a url, look there :p

Great to know. I thought it could be something like that, so I am very glad to have your input & validation of that

you folks rock

if the cause is not obvious, I would recommend logging the result of url.parse on whatever URL you are trying to access in the code that fails, and seeing whether the hostname is what you expected

it can of course also just be that the domain really doesn't exist :)

Ugh. I think I see what is happening. My company/team uses OOP. So, in this project, there's the real world object, which connects to an enterprise app. Then, we test that object... but using the sandbox account instead of the live production account. I extracted some functions from the main real world object, and those functions rely on the main real world object... yet those functions are also used in the test objec

t

This is why I don't like OOP -- things are tied together

can someone help me understand something, so, I have this version of node installed -> nodejs-14.15.4_1 <- however, I also have nvm installed. SO when I open bash and type node -v I get v12.19.0, when I do it in any other terminal (fish or elvish for example) I get v10.10.0... how? evlish and fish shouldnt have nvm working on them anyway. where is the node version determined?

yaalon: that sounds like "bad OOP", not "OOP" :P

Yeah... I think you're right. OOP isn't necessarily the culprit. I just prefer functional programming so I can keep clear separations between __everything__ and not make things dependent mounting functions on objects

Kingsy: if you use fish/elvish' mechanism for determining where a certain binary/command comes from, what do you get?

in Bash it'd be `type -a` or `which`

ok, I have a stupid question... how can I save the result of a sql query to an object, or is that a bad idea

moment

joepie91: /usr/local/bin/node

Camilo: it's already an object

and that gives me v10.10.0

Kingsy: have you ever used `n`?

maybe before trying nvm

what do you mean?

nvm isnt even available on the prompt.

weirdly, according to me package manager, it installs it here -> /usr/bin/node <- so how its going for /usr/local/bin/node first, is the mystery

Presumably to have the result you described, you have Node installed at least three different ways.

samsch: exactly.. which is just mental.. :D I mean I do have nvm. which is great, but I don't want it to be used outside of bash

If the nvm Node is being used, `which node` (or equivalent) will show a path with nvm in it, from your user folder. My guess is that nvm in bash just needs you to add the nvm .bashrc stuff.

yeah Ih ave done that. so nvm works in bash, but my default terminal is elvish, so I would expect when I run node -v there it would be the OS version of node, not nvm related.

If you did the `nvm use system` thing or whatever though, that could also let you see the non-nvm path to node in bash.

That sounds like it's already doing that.

But that you have two OS versions of Node.

elvish isnt ANSII you see, so I cant interact with it within this shell

samsch: hmmm wow. perhaps!

I don't see how I can, unless it was installed via npm or something.

I think that was the reason for joepie91's question about having used `n` in the past, because it can have that result.

alright, let me seeeee

holy crap thats it!!!

right, I mentioned `n` because it messes with globally-installed Node, and is usually the cause for this specific problem :)

right this is getting removed then

and a lot of people try `n` before they try `nvm` and then forgot they ever used `n`, and only discover later that `n` messed with their global environment

when they try to use a system Node

`n` acts too much like malware :/

It's like one of those "technically correct" solutions that's just wrong.

well there you go. I am back on v14.15.4

I never ever would have remembered that.

thanks so much joepie91 and samsch

good to see it solved :)

sad trombone noises

clearly I'm a bit slow because I still can't seem to access my sql result object outside of sql callback function

Camilo, You'll need to do your work inside the callback. (And hopefully in a Promise .then() callback rather than directly passed)

samsch: thanks, I am working towards that now.

banging my head against the wall because I broke my sql query somehow

do you know if it is ok to nest queries inside of callback functions of other queries?

You probably don't want to. Use promises, let the caller get back the data or the transformed data in its own .then(), pass the data to another function from there if you have to

break it down into single functions that each to one thing

!paste @ Camilo

Camilo: Please paste your code to a site like gist.github.com or dpaste.org (NOT pastebin.com or hastebin), or create a running example: see !fiddle. See also !badpastebin !hastebin !nopaste and !mcve

also if you gist the code we can give you more specific advice

ok thanks will add some comments and upload, very kind of you!

Does adding OOP keywords-- Class, Public, Private, Static... feel like nodejs is getting turned into java/an OOP language to anyone else?

Dunno, I just don't use any of them :)

I guess it just allows an OOP way of using nodejs... but it's a bit weird to me... why not just use an OOP based language to use OOP stuff?

zsoc: glad I am not alone on that. but now I am thrown into an organization & team which apparently all use OOP and want to sort of "do things as they've always been done here"

yaalon: yeah, I'm not a fan of the class scope creep either

struggling with the "todo" part, imo the easiest solution is to make 3 different SQL quiries, but maybe thats not a goo idea

yaalon: it's all basically unnecessary because lexical scoping pretty much already does all of this?

but people insist on using their damn classes :(

yaalon: JS is not adding "public" or "private" keywords.

yaalon: that's part of typescript.

the `class` keyword, while badly named is an eminently superior way of doing inheritance in JS than previous alternatives. it's just that inheritance itself sucks.

And what *would* have made sense is to deprecate doing inheritance at all

but noooo

that would never have been even remotely possible

I worked with Clojure a bit about 2 years ago. And my goodness, separating functions from data seems to frikkin intuitive. Having data objects & functions in javascript seems so much easier to grok than tying functions and data into objects... because then you're stuck with those objects-- it's hard to break them down once they get built up.

so* intuitive

yaalon: that's accurate

yaalon: but that's not an argument for making inheritance codebases less terrible

So much simpler than using all these funky keywords...

argh. lol

*for not making them less terrible

yaalon: again tho there's no public or private keywords

in JS, everything reachable is public, everything unreachable isn't. class private fields use `#` to indicate that the fields aren't reachable from outside the class body

ljharb: thanks! glad to have that clarification. I was confused because I was reading about them here-- sitepoint.com/javascript-private-class-fields

And then this mentions "public class fields" developer.mozilla.org/en-US/docs/We…ference/Classes/Public_class_fields

but I see that it's a new/experimental feature

Lol... Ugh I just don't want to see people adding [what seems to be] Ruby & Java concepts into NodeJS and turning it into those languages lol

yaalon: public class fields are automatically public, like all object/class properties

ljharb: wasn't JS adding public/private *syntax* though?

yaalon: it may help you to know that private class fields are unique out of every language on the planet

Awesome ljharb. It's so awesome that the language developers such as yourself & other folks are here to chat with

joepie91: not really no

yeah that

public fields are just normal field syntax

it's not the public/private keywords. public fields are just normal per-class properties

@Camilo: sorry i got pulled into a meeting, i can look at it in a few if you stick around

but JS is the only language i know of that has *actually* private class fields.

yaalon: ruby and java both let you break encapsulation whenever you like

brb, going to a cafe to get some grub. Ill. be back in 10 minutes. and ill try to find the log of this convo so I can keep up

@zsoc: thanks, I'm just reading up on promises at the moment, appreciate being steered in the right direction!

in dockerfile, do you have CMD to run npm start? or just run node ... ?

saml, often `npm start` will run the development server

so either a separate script or `node …`, in the common case

it's a bit awkward. this project does npm start in docker because prestart does sequelize db:migration.. hrm what disaster can happen when multipel containers run db migration at start?

you should run the migration as a separate container

yeah.. not sure why app does migration

it might be the same docker image, but with a different command

maybe it's easier in docker-compose to synchronize migration && running node app

if migration were in different container, node app container still needs to wait for migration container to complete

I've only worked with kubernetes for container orchestration, and there I can run a "job" and wait for it to complete

you could do something similar with ssh + docker commands. Not sure about other orchestration tools

Is there any well-supported library to make MySQL queries that supports promises?

AAA_awright_: doesn't mysql2 do that?

Isn't that an unofficial library? I ask because I just ran into an issue with the connection pool that pretty much makes it unusable

AAA_awright: all libraries are "unofficial", AFAIK :)

have you filed a bug?

Yeah, and regression

in that case, use an older version?

And I mean, more or less, some more on the more-unofficial side

I'm trying to upgrade a project from node12 to node14. It's an express webserver. Everything seems to work except in one spot where I pipe a file stream to a response. This works under node12 but fails under node14. I'm totally lost as to why. If I manually listen for the data and end events from the stream and write to the response it works as well. It's only when I try to pipe directly to the response that it fails. Can

anyone think of any node changes that could have caused this? Or even a place to start looking for what I'm doing wrong? I've been chasing this for like two days without much success (other than finding that manually writing worked)

tu: can you pastebin/gist the code?

The express example in the answer here stackoverflow.com/questions/5823722…/how-to-serve-an-image-using-nodejs

It's in an app.use instead of app.get but it's basically that simple.

Also tried using the npm send module (which expressjs uses for serve static) and it acts the samw

tu: can you gist/pastebin your *actual* code? :)

problems are often in unexpected subtle places

Ain't gonna help but here gist.github.com/theunforgiven/40bc7b02b86dd66aa40e7866bdbe5b84

All the commented out code is other things I've been trying

70-78 works. As does 57/58

tu: when you say it 'fails', what do you mean exactly?

When the browser requests it, it just hangs and never responds and is eventually r

Timed out on the browser side

Node sees the request successfully complete with a 0 content length

Also keep in mind this works in node12 exactly as is and only starts failing under node 15

Er node 14

very odd

it smells like a Node bug but it seems too-fundamental functionality to be broken like that

are you on latest 14.x?

tu: still fails in node 15?

Yes. Downloaded it yesterday

it might be an intentional breaking change in node 14.

Have not tried node15 that was a typo.

It's weird though because the same route works fine when called in the browser

So as much as I'd love to blame it on node (because I'm so sick of debugging this at this point heh) I remain unsure

tu: "works fine when called in the browser"?

The route works fine if I hit manually in a browser

as opposed to?

in the browser it likely uses a streams polyfill, which may not have updated to account for node 14

It only seems to die when my browser app requests it

ah. are the headers different in those two scenarios?

By headers do you mean request or response?

tu: request

Hrm. I don't think content type stuff because I can trace into the server code and see it reach the path where it opens the file and starts to pipe it back

Both paths definitely lead to it trying to send the file. Just manually the browser downloads it, where as the js client times out after 10 seconds

Note the browser response happens instantly so it's not just a slow route

tu: adjusted suspicion: there's something weird about how your JS consumes the response (eg. the rate at which it does so), the interpretation of which on the Node.js side was changed from 12 -> 14

I think the slow HTTP request attack mitigations were longer ago than that?

The only weird thing is toggling the node version back fixes it

tu: I would definitely read the 13.0.0 and 14.0.0 changelogs to look for anything that changed in the http implementation, especially surrounding timeout handling, attack mitigations, etc.

I don't really own the code that is trying to do the download. Lemme see if i can find it and see if it's doing anything weird

that also helps :)

From the minified version it looked like a standard request and settimeout handler.

I'm so baffled as to how a now change could even cause this.

Especially when I can manually stream the file out / read it synchronously and it starts working

I'm honestly not sure either

Fwiw the browser side is basically just.

Er basically just hls.js so it's not just some random code it should be a reasonably done heh

Is it normal to have a readStream.pipe on both line 105 and 111? What's the behavior when doing that?

Oh no. That's just cruft from the copy paste.

Ok, nvm then. That's what I assumed.

I updated the gist and fixed that and added one other tid bit. I added a handler for the end event on the stream and curiously enough it never fires

So I think part of the problem is something around the stream never being closed or something because it never seems to fire events like it was

Ooohhh. I think I am starting to narrow something down. My code works with the last version of 13.14.0 so something after that is causing my issues. Now I can narrow it down at least

tu: it's odd, because end-of-streams *should* be passed through with a pipe

tu: out of curiosity, what happens when you use streams.pipeline?

One sec will try that

Stream pipeline fails the same way and my callback for it is never actually called.

another stupid question: if I put my functions in a module, how do I access the objects in my main program? Do i need to also include them in the modules?

So it works with the last version of 13.x and starts failing in 14.0.0

Camilo: sorry again, today turned into a mess lol. So one way people do it is by making their modules parametric (return a function that takes parameters and returns the module)

Camilo: So like if you have a module that needs your db client, you can do `const createFoo = require('./my-foo'); const foo = createFoo(db); Or maybe pass it an object with all stateful deps instead of just the one individually (So it's extensible)

Camilo: this is an example of that pattern with an express router, but ofc you can use it with anything: gist.github.com/jkantr/b3706357a831108579bb0f152b7768cb

Hello, irc.mesgul.net:8337 How can I solve the character error ?

Hello, irc.mesgul.net:8337 How can I solve the character error ?

hi

are there any reporting engine in Nodejs?

Apparently disabling keep alive fixes my weird pipe streaming issue.