metron-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Tamás Fodor <ftamas.m...@gmail.com>
Subject Re: [DISCUSS] Switching to a better alternative of Pikaday.js
Date Wed, 17 Oct 2018 08:26:54 GMT
Organizing the codebase as a monorepo could be a good idea to address it.

There's a popular and very powerful tool called Lerna
<https://github.com/lerna/lerna> to maintain a monorepo:

"Splitting up large codebases into separate independently versioned
packages is extremely useful for code sharing. However, making changes
across many repositories is messy and difficult to track, and testing
across repositories gets complicated really fast.
To solve these (and many other) problems, some projects will organize their
codebases into multi-package repositories (sometimes called monorepos).
Projects like Babel, React, Angular, Ember, Meteor, Jest, and many others
develop all of their packages within a single repository.
Lerna is a tool that optimizes the workflow around managing multi-package
repositories with git and npm.
Lerna can also reduce the time and space requirements for numerous copies
of packages in development and build environments - normally a downside of
dividing a project into many separate NPM package."

To me, it seems to be a good approach to deal with what we have in the
metron-interface folder. It forces you to write independent and focused
packages regarding the single responsibility responsible. You can easily
share them across the alerts and the management UI. I highly recommend you
to go through the docs and think about it.

Btw, shouldn't we register an organization on npm to publish our packages?
@hortonworks is already taken. Does it belong to us? Do you know who's got
permission to publish packages under it? Can we also have the rights to
publish?

Cheers,
Tamas

On Tue, Oct 16, 2018 at 6:50 PM Michael Miklavcic <
michael.miklavcic@gmail.com> wrote:

> I'm still +1 to this approach. Thanks guys!
>
> To Tibor's other point - the management UI is currently a completely
> separate service, as I understand it. If there is infrastructure you think
> is shareable while still allowing them to be independently deployed, then I
> think it would be desirable to make improvements there also. My only
> caution would be that there are instances where early architectural
> duplication seems wasteful, but is really just a short-term illusion
> because of code maturity. ie, there are instances where you absolutely
> don't want to tie architectural components together even if it seems like
> you're duplicating code. I don't believe this fits that concern, but I
> thought it was worth mentioning. To that end, I wouldn't have your progress
> on the alerts UI improvements be hampered by the management UI. I think we
> can take the lessons learned in the Alerts UI and apply them to the
> management UI when it makes sense to.
>
> Best,
> Mike
>
>
> On Tue, Oct 16, 2018 at 2:21 AM Tibor Meller <tibor.meller@gmail.com>
> wrote:
>
> > Thanks, Tamas! The plan you described seems a good approach to me.
> > Let's wait another day before we create Jira tickets from these steps to
> > make sure no one else has other concerns.
> >
> > One more question: how much of these changes could be applicable to the
> > Management UI?
> > It would be great plus to see those two UI getting closer to each other
> > with the underlying technologies instead of shifting away.
> >
> > On Tue, Oct 16, 2018 at 9:37 AM Tamás Fodor <ftamas.mail@gmail.com>
> wrote:
> >
> > > I'm agreeing with moving forward with Ng Bootstrap. In order to do
> that,
> > I
> > > think it would be a good start to refactor those components which use
> the
> > > obsolete jquery based vesrion of bootstrap. Shane has already started
> it
> > by
> > > refactoring the Modal dialog in the management UI (We also have this
> > > component in the alerts UI). Once we've succeeded that, we can remove
> > > (hopefully) jQuery form the dependency list.
> > > As the second step, we should migrate to Ng bootstrap. I'm assuming
> that
> > if
> > > the jquery based Boostrap is replaced with Ng bootstrap, we're still
> good
> > > since we're using only classes from the bootstrap CSS which is shared
> > > across jQuery bootstrap and Ng bootstrap.
> > > If everything is good and nothing is broken we can start working on the
> > new
> > > date/time picker component based on Ng bootstrap and then we can get
> rid
> > of
> > > Pikaday.
> > > Once Pikaday is completely eliminated from the system, we can be sure
> > that
> > > the PR about eliminating moment.js is ready to go.
> > >
> > > Cheers,
> > > Tamas
> > >
> > > On Mon, Oct 15, 2018 at 8:36 PM Michael Miklavcic <
> > > michael.miklavcic@gmail.com> wrote:
> > >
> > > > I like point #3 from Tibor - you can pick and choose how you compose
> > the
> > > > date/time pickers. It would be nice to not have times required as a
> > > > dropdown at some point or another, or at the very least something we
> > can
> > > > customize differently to our liking :)
> > > >
> > > > Per the comments from Tamas, Shane, and Tibor, is there any reason we
> > > > wouldn't want to move forward with the Angular port of Bootstrap, NG
> > > > Bootstrap? Per your arguments for it, this sounds like the right path
> > > > forward to me. I'm +1 on this approach provided someone doesn't come
> > back
> > > > with a "well, there's only this problem..."
> > > >
> > > > Best,
> > > > Mike
> > > >
> > > > On Mon, Oct 15, 2018 at 7:39 AM Shane Ardell <
> shane.m.ardell@gmail.com
> > >
> > > > wrote:
> > > >
> > > > > We are definitely on the same page.
> > > > >
> > > > > On Mon, Oct 15, 2018 at 2:06 PM Tibor Meller <
> tibor.meller@gmail.com
> > >
> > > > > wrote:
> > > > >
> > > > > > @Shane It seems like we're agreed on this. :D
> > > > > >
> > > > > > On Mon, Oct 15, 2018 at 1:18 PM Tibor Meller <
> > tibor.meller@gmail.com
> > > >
> > > > > > wrote:
> > > > > >
> > > > > > > Hi Guys,
> > > > > > >
> > > > > > > I think we could consider moving to ng bootstrap. It solves
> most
> > of
> > > > our
> > > > > > > problems and makes our code base cleaner easier to maintain.
> > > > > > >
> > > > > > > Here are some benefits I see:
> > > > > > > 1. we could eliminate jQuery from the code base
> > > > > > > Currently, we're using bootstrap but an implementation
based on
> > > > jQuery.
> > > > > > > Angular and jQuery doesn't build to live together.
> > > > > > > 2. NgBootstrap made to being used with Angular
> > > > > > > It uses Angular instead of hacking the rendering/dom
> manipulation
> > > > with
> > > > > > > jQuery.
> > > > > > > 3. It contains a date and a time picker.
> > > > > > > It's easy to combine to a datetime picker.
> > > > > > > 4. No dependencies.
> > > > > > > By changing currently used bootstrap to NgBootstrap we
could
> > clear
> > > > > > jquery,
> > > > > > > moment, pickaday, pickaday-time libraries from our
> dependencies.
> > > > > > >
> > > > > > > Another huge advantage of NgBootstrap is that we don't
have to
> > > > rewrite
> > > > > > > anything we don't want to. Our UI already uses Bootstrap.
> > > > > > >
> > > > > > > What do you guys think?
> > > > > > >
> > > > > > > On Wed, Oct 10, 2018 at 4:19 PM Nick Allen <nick@nickallen.org
> >
> > > > wrote:
> > > > > > >
> > > > > > >> > Before making a decision on what's next, I'd to
ask you a
> > > > question.
> > > > > Is
> > > > > > >> it really
> > > > > > >> a priority and is it really worth the effort to touch
our
> > > currently
> > > > > used
> > > > > > >> date picker component to get ~15% reduction in the
bundle size
> > by
> > > > > > removing
> > > > > > >> moment?
> > > > > > >>
> > > > > > >> As an aside, I think there is a greater benefit here
too.  We
> > need
> > > > to
> > > > > > make
> > > > > > >> a conscious effort to identify libraries that we are
using
> that
> > > are
> > > > > > >> deprecated, lack community support, and are unlikely
to be
> > > > maintained
> > > > > > and
> > > > > > >> updated for security vulnerabilities.  We need to actively
> > > identify
> > > > > and
> > > > > > >> replace those.
> > > > > > >>
> > > > > > >>
> > > > > > >>
> > > > > > >>
> > > > > > >>
> > > > > > >> On Wed, Oct 10, 2018 at 9:33 AM Tamás Fodor <
> > > ftamas.mail@gmail.com>
> > > > > > >> wrote:
> > > > > > >>
> > > > > > >> > I'd like to open a discussion about switching
to a new date
> > > picker
> > > > > > >> library
> > > > > > >> > in the Metron Alerts UI regarding to the following:
> > > > > > >> >
> > > > > > >> >
> > > > > > >> >
> > > > > > >>
> > > > > >
> > > > >
> > > >
> > >
> >
> https://lists.apache.org/thread.html/2e4fafa4256ce14ebcd4433420974e24962884204418ade51f0e3bfb@%3Cdev.metron.apache.org%3E
> > > > > > >> >
> > > > > > >> >
> > > https://github.com/apache/metron/pull/1219#discussion_r223733562
> > > > > > >> >
> > > > > > >> > A week ago, I opened a PR about removing moment.js
from the
> > code
> > > > > base
> > > > > > to
> > > > > > >> > decrease the size of the production javascript
bundle. I
> could
> > > > > achieve
> > > > > > >> 15%
> > > > > > >> > loss in the final bundle size which is admittedly
not a game
> > > > changer
> > > > > > but
> > > > > > >> > still. Not to mention if we want to heavily rely
on date
> > > > manipulator
> > > > > > >> > functions in the future it's better to get rid
of it at this
> > > early
> > > > > > >> stage.
> > > > > > >> > Go here <https://github.com/apache/metron/pull/1219>
to
> read
> > > more
> > > > > > about
> > > > > > >> > the
> > > > > > >> > background and the results. I tried to provide
as many
> details
> > > as
> > > > I
> > > > > > >> could.
> > > > > > >> >
> > > > > > >> > So far, so good. But then I stumbled upon an obstacle,
> Pikaday
> > > > > > >> > <https://github.com/dbushell/Pikaday>.
> > > > > > >> >
> > > > > > >> > Before going further, let me thank Tibor Meller,
Michael
> > > Miklavcic
> > > > > and
> > > > > > >> > Nicholas Allen for taking their time to go through
my
> proposal
> > > to
> > > > > deal
> > > > > > >> with
> > > > > > >> > the aforementioned issue. At the end, we agreed
on basically
> > not
> > > > > going
> > > > > > >> with
> > > > > > >> > my temporary solution that intended to solve the
related
> > > problems
> > > > of
> > > > > > >> > Pikaday and we'd rather like to find and change
for a better
> > > > > > >> alternative.
> > > > > > >> >
> > > > > > >> > To be fair, Pikaday is a pretty good date picker
module, its
> > > only
> > > > > > >> problem
> > > > > > >> > is the moment dependency if it's installed via
npm. But
> other
> > > than
> > > > > > >> that, it
> > > > > > >> > functions perfectly. Zero dependencies, small,
etc. Long
> story
> > > > > short,
> > > > > > >> it's
> > > > > > >> > good for us unless we want to get rid of moment.js.
> > > > > > >> >
> > > > > > >> > Before making a decision on what's next, I'd to
ask you a
> > > > question.
> > > > > Is
> > > > > > >> it
> > > > > > >> > really a priority and is it really worth the effort
to touch
> > our
> > > > > > >> currently
> > > > > > >> > used date picker component to get ~15% reduction
in the
> bundle
> > > > size
> > > > > by
> > > > > > >> > removing moment?
> > > > > > >> >
> > > > > > >> > I'm asking it because if we want to do so, considering
that
> > > it's a
> > > > > > huge
> > > > > > >> > topic, the following questions might come up:
> > > > > > >> >
> > > > > > >> > *A: What component do we want to use instead of
Pikaday?*
> > > > > > >> >
> > > > > > >> > I'm not satisfied with the alternative individual
solutions
> > out
> > > > > there
> > > > > > on
> > > > > > >> > npm. I'd rather pick a component library like
the angular
> port
> > > of
> > > > > > >> Bootstrap
> > > > > > >> > <https://ng-bootstrap.github.io/#/home>
or the angular
> > material
> > > > > > library
> > > > > > >> > <https://material.angular.io/>. Both of
them have a date
> > picker
> > > > > > >> component
> > > > > > >> > and many other components to rely on and reuse
throughout
> the
> > > > Metron
> > > > > > >> app.
> > > > > > >> >
> > > > > > >> > *B: What component library do we want to use?*
> > > > > > >> >
> > > > > > >> > Introducing a new component library requires a
lot of
> research
> > > and
> > > > > > there
> > > > > > >> > are many things we have to agreed on. Since it's
a long term
> > > plan
> > > > > > >> because
> > > > > > >> > it would be great to use it consistently instead
of picking
> a
> > > new
> > > > > one
> > > > > > a
> > > > > > >> few
> > > > > > >> > months later just because we chose wrongly.
> > > > > > >> >
> > > > > > >> > *C: What about the jQuery version of Bootstrap?*
> > > > > > >> >
> > > > > > >> > So basically we already have a component library
and we
> still
> > > use
> > > > it
> > > > > > but
> > > > > > >> > we're also planning to replace it with another
or the
> angular
> > > port
> > > > > at
> > > > > > >> least
> > > > > > >> > to get the most out of the angular rendering engine.
Since
> it
> > > uses
> > > > > > >> jquery,
> > > > > > >> > it's much less performant than a port written
in Angular.
> > > > > > >> > And I think it's a bad idea to introduce a new
one and use
> > > > multiple
> > > > > > >> > component libraries within one project.
> > > > > > >> > We can also pick the date picker component from
the jQuery
> > > > Bootstrap
> > > > > > >> but,
> > > > > > >> > again, it's not as efficient as the angular port
so it seems
> > to
> > > be
> > > > > > >> > beneficial to replace it with something else.
> > > > > > >> >
> > > > > > >> > What do you think, guys?
> > > > > > >> >
> > > > > > >> > Thanks,
> > > > > > >> > Tamas
> > > > > > >> >
> > > > > > >>
> > > > > > >
> > > > > >
> > > > >
> > > >
> > >
> >
>

Mime
  • Unnamed multipart/alternative (inline, None, 0 bytes)
View raw message