Any plan to replace "request" and "moment"?

karsten13

@0m4r we had similar discussions already here

swvalenti

How can I tell if a module will be compatible prior to the fall upgrade of 2.17.0? I don’t want to upgrade and half of my modules break.

0m4r

@swvalenti if your module uses “request” to issue HTTP calls then it is.
so, if you import it in your node_helper.js (or an alike node pice of code).

i.e.
const request = require('request');
or
import request from 'request'

on way to preserve it is to add “request” as a dependency to your module, but your end-users will need to install them on their own.
To add it to your module, from its folder you can do:
npm install request ; this will update your package.json file (and package-lock.json) - commit the change to your git repo (this is most likely a breaking change and would require a major version increase - if you version your code!)

Once this is done, your existing (and new) end-user will need to update their copy of your module and run npm install on their own to download request and enable them to use it.

swvalenti

@0m4r said in Any plan to replace "request" and "moment"?:

const request = require(‘request’);

thank you for the explanation I understand better now

karsten13

@swvalenti one addition: Only modules are affected which rely on that request is intalled in core mm (which is a dirty hack of the module developer), modules which install request on their own are not!

sdetweil

@karsten13 said in Any plan to replace "request" and "moment"?:

which is a dirty hack of the module developer

no… don’t say it that way…

lack of experience on the part of the developer…

it just worked, so they didn’t know to ask if there was a more correct way…

I had that problem early on in my nodejs development too…

karsten13

@sdetweil doing this in some of my own modules so I’m allowed to say this ;)

sdetweil

@karsten13 doing it ON PURPOSE is a hack!!

0m4r

@sdetweil @sdetweil I am not convinced about both of your points of view. I mean technically yes.

But… looking into the modules themselves, they are solely made to work with the MagicMirror code base (99% of the cases)
What is the point to bloat the size of the installation with the same node module installed over and over again? when on, “global” can be used and relied upon? (well, yeah, I agree, “this situation” is one reason for it!). But why trade-off convenience over complexity?

I see it as a part of the Magic Mirror ecosystem, provide some common “code” to be used by everyone to both ease the development and the set-up of it.

As mentioned in the other thread, old legacy modules will probably stop working - forking them or re-doing them will fix them. This not only pushes forward the advancement of the whole MagicMirror tech stack, but it may strengthen the community and really push it to hold together and upgrade, modernize the code of the modules.

bottom line, nobody says it is going to be easy, but it is a big opportunity.

sdetweil

@0m4r said in Any plan to replace "request" and "moment"?:

What is the point to bloat the size of the installation with the same node module installed over and over again

it is NOT installed over and over

npm install start with the module folder and says

is it here?   
   no   
     look in the parent folders, 1 by 1, looking for a node_modules folder
       is it there?
          yes, stop
   yes, 
     stop.

so it will only be installed once…

in the near term with all the old modules ‘assuming’ is was installed,
it’ll have to be installed in the MM/node_modules (but only once)

0m4r

@sdetweil yeah, I guess, I was not clear enough. And I think we are saying the same thing.

Installing “request” in the MM root folder solves the problem, and this is how I do run my magic mirror from the development branch at this very moment.

What I mean is that if the module is removed from the “MM/node_modules” folder (the last “parent” level for all modules), every single module will most likely need to install it on its own node_modules folder. This is where I meant it may blow up the size of the MM installation (modules included!)

I hope this clarifies my prev post :)

sdetweil

@0m4r said in Any plan to replace "request" and "moment"?:

What I mean is that if the module is removed from the “MM/node_modules” folder (the last “parent” level for all modules), every single module will most likely need to install it on its own node_modules folder. This is where I meant it may blow up the size of the MM installation (modules included!)

yeh, you have NO idea how big other stuff is already…

but, this is the design… there is virtual memory, and paging and … to offset these

now, remember, we are NOT making an installable bundle from all this, just a runtime layout

BKeyport

God, the alternatives to moment are head spinning. I don’t understand 'em, and can’t even get basic functions to work right within jsfiddle.net (Example: Date1 <= Date2 crashes jsfiddle)

For now, I’ve added package.json to both modules, and if needed, will update my install instructions.

MMRIZE

For the moment, maybe luxon or date-fns could be the alternative. (And there are several other modules also.)

However, After releasing TC39’s new proposal (currently stage 3. https://tc39.es/ecma402/), Browsers and NodeJS will have the native features of handling dates. So I would recommend waiting.

At this moment, native implementing moment’s features is not easy. Though Intl could handle many aspects of moment already, the unsolved big issue is, current Date and Intl features are still not perfect for localization, like weekOfYear, startdayOfWeek and so on. (Those will be solved by ECMA402)

0m4r

@mmrize agree to support native browser APIs, but I do also believe we cannot assume everyone is running an up-to-date system for their magic mirror.

One idea could be to provide a wrapper module to abstract the “behind the scene” tool to manipulate dates, but it may turn out to be quite of an effort and limiting the module developer.

cowboysdude

@mmrize Well we all know how ‘browsers supporting’ all work out LOL

Some do…some don’t.

sdetweil

@cowboysdude AND for mm, the calendar date work is done OUTSIDE the browser in the node helper. soooooo, still need something

MMRIZE

@cowboysdude @sdetweil
Fortunately, We need to consider only ‘Chromium’ and ‘node’ at most. (Firefox already implements almost-complete Intl features. For Safari… who cares.)
And in this case, Intl (and the Temporal in the future) suggests the STANDARD WAY how to handle date/time for L10N/I18N in Javascript environments(includes browsers and nodeJS). So I believe it has worthy looking inside.

Once (yesterday), I’ve tried to rewrite clock module and calendar module without momentjs or any other 3rd party library, only with Date and Intl.

The start was not so bad. I could rewrite many features with only Date and Intl.
Where I stuck was the showWeek config feature. It has to show the ordinal number of weeks of the year. To calculate ISOweek is not difficult. But in some areas, like the US. they never use the ISO system. damn!. To calculate the conventional US week number, additional pieces of information are needed. (read this.) Current Intl has not yet that feature. I’m waiting for Temporal and enhanced Intl releasing.

One solution might be to obtain additional information needed for conventional calculation from the user by configuration. At this point, I stopped rewriting. It needs to change config.js and that was not my first intention.

A usual solution might be to use 3rd party library like momentjs or luxon. It is easier.

However it’s good to read this from the momentjs (https://momentjs.com/docs/#/-project-status/future/) I agree to him, In the near future, native JS features Date, Intl, Temporal will be a standard.

sdetweil

@mmrize i don’t see whats so hard on week of the year.

what day of the week is 1/1, subtract that from 7, and then that is the start of the weekly cycle.
til you run out of days to count.

we don’t use it normally, but its still the week of the year. has nothing to do with month. that is a different scheme

my science teacher in high school had a great topic, measurement…
you can measure anything with anything… just record what the tool was

we were required to measure everything in broomstick handles, broomstick #1 in the lab to be precise.

we could subdivide broomstick #1 however we wanted to provide more accuracy, as long as we wrote down the algorithm. (divide by 3, 6, 10?.. , 1000 ml beaker height, whatever)

the last week doesn’t run over into a new year, it ENDS when the year ends.
just so happens that week #1 of the NEXT year is the same logical week, but who cares, as we started counting from 1 again.

MMRIZE

@sdetweil said in Any plan to replace "request" and "moment"?:

what day of the week is 1/1, subtract that from 7, and then that is the start of the weekly cycle.

ISOWeek is easy. Conventional week is the problem.

For example, in the United States, Sunday is the first day of the week. The week with January 1st in it is the first week of the year.

In France, Monday is the first day of the week, and the week with January 4th is the first week of the year. It means January 1st could be the last week of the last year sometimes. And Jan. 10th could be the first week in certain year.

And in other countries, there is also their own rule.

This might be important in some European countries, because they use week number in real life (“my vacation ends 23. Weeks“)