►
From YouTube: JupyterLab Weekly Dev Meeting, Friday April 7, 2017
Description
Meeting of the JupyterLab development team.
Meeting Notes: https://paper.dropbox.com/doc/JupyterLabNotebook-Weekly-Meetings-rIhXeFWYRgCiCFKiz2gv4
Meeting Note Archive: https://jupyter.hackpad.com/JupyterLabNotebook-Weekly-Meetings-UUJ3gIQ3iBS
A
Welcome
Internet
to
the
Jupiter
lamb,
/
notebook,
dev
meeting
Friday
April.
Seventh,
we've
just
switched
to
Dropbox
paper
to
try
taking
our
notes,
taking
a
lead
from
the
normal
Jupiter
dev
meetings
and
first
up
looks
like
we
have
my
report
aight
I
legit.
So
we've
also
switched
the
notes
to
more
project-based
rather
than
person
based.
A
So
I
guess
first
up
is
I
pie.
Widgets
we've
been
pushing
really
really
hard
to
try
to
get
a
widget
7
out
I've
held
off
on
doing
a
duper
life
extension
because
I
saw
you
were
working
on
the
Jupiter
geo
JSON
extension.
I
thought
I'll.
Let
you
figure
out
all
the
dark
corners
of
actually
getting
a
third
party
extension
working
and
and
we've
had
plenty
of
work
and
I
five
widgets.
A
The
upcoming
out
hi
widget
7
release
says
that
if
I
have
no
shortage
of
things
to
occupy
me
all
night
long,
instead
of
sleeping
so
we've
been
working
a
lot
on
cleaning
up
sinking
bugs
and
taking
a
fresh
look
at
a
lot
of
the
lower
level
architecture
as
well
as
cleaning
up
a
few
regressions,
and
things
like
that.
That
was
noticed
from
Ridgid
six.
So
we
hope
to
have
legit
seven
I'm,
hoping
by
the
end
of
april
out
and
hopefully
helpful
before
that.
B
That
was
mostly
down
the
path
of
revamping
the
extension
builder
do
do
late,
building
on
the
user's
machine,
as
extensions
are
added
or
removed
at
well.
Removing
them
could
be
just
disabling,
but
at
least
when
you
add,
you
would
need
to
refill,
and
the
reasons
for
this
were
that
we
couldn't
effectively
merge
built
artifacts
and
Andy
duped
them
at
the
same
time,
because
those
artifacts
wouldn't
sir,
include
all
the
files
needed.
B
So
the
thought
was
to
bundle
the
entire
node
modules,
Creed
merge,
said
trees
and
then
do
a
funnel
web
pack
built
for
the
the
artifact
that
actually
gets
loaded
onto
the
page.
All
that
is
working,
except
for
the
fact
that
the
step
of
duplicating
these
trees,
so
we
just
take
a
node
modules
directory
and
throw
a
number
of
trees
at
it.
It
works
unless
there's
a
conflict,
in
which
case
the
the
standard
am
PMD,
Duke
wasn't
rectifying
the
complex,
even
though
there
was
no.
B
There
was
no
conflicting
library
that
would
cause
it
not
to
hoist
it,
but
we
saw
the
case
where
and
it's
most
likely
because
it
wasn't
installed
to
wait.
Npm
expected
it
to
be
installed.
It
wasn't
in
a
tree
that
it
knew
has
a
recovered
from
to
do
this
horse
thing.
So
if
the
package
is
exactly
the
same,
it'll
hosted,
but
if
it's
different
in
any
way,
even
though
it's
compatible-
it's
not
hosting
it.
B
So
there
are
those
precedents
for
using
an
MPN
cash
local
cache.
That's
just
we
just
came
to
this
realization
rep
for
being
so
that
seems
to
be
the
same
as
path
forward
unless
there's
an
option
where
we
don't
want
to
I
hate
a
date
to
do
this,
we
should
like
English
you
Lisa
for
that
possibility
before
we
declare
this
the
winner,
because
there
is
the
we
could
take
attack
of
saying
no,
you
NPM
install
from
the
internet
period,
but
I
don't
think
we
want
to
just
go
down
that
path
and
then
find
out.
B
B
So
what
did
what
I
did
to
try
to
fool?
It?
Was
that
told
em
PMD
dubai
actually
edited
the
DD
code
to
say:
don't
look
at
that
metadata,
that's
been
adding
and
when
that,
when
it
did
that
it
properly
hosted
it
and
that's
so,
but
the
problem
was,
it
was
adding
metadata
assuming
it
was
already
in
a
valid
node
modules
tree
which
wasn't
because
we
had
just
thrown
stuff
in
there.
B
So
if
we,
if
we
hacked
nvmd
Duke,
then
it
works,
but
does
it
really
work
as
a
question
at
that
point,
because
presumably
that
metadata
was
there
for
a
reason,
so
we've
yet
we
could
hack
around
the
solution.
But
also
of
note
is
that
mpm
doesn't,
unless
they've
broken
out
into
a
separate
package
under
the
NPM
org.
It's
not
a
public
programmatic,
API,
so
npm
dee
doop
is
not
a
package.
B
B
So
what
we
came
to
the
stance
was
no
you.
No
one
else
can
provide
loaders,
we
have
a
set
of
loaders
and
if
you
need
something
different,
you
have
to
transform
your
code
first
to
match
our
set
of
things
that
we
can
pack,
and
that
includes
es5,
JavaScript
CSS
and
some
image
and
image
files,
JSON,
etc.
This
is
a
list
of
file
types
that
we
can
support.
So
if
you
have
something
like
less
files,
they
would
have
to
be
trans
filed
before
putting
them
into
a
Google.
At
extension,
that
would
listen
soon.
Would
it
be?
B
B
Where
we
thought
about
that,
because
the
problem
is,
it
is
so
you're
providing
a
loader
stay,
for
example,
a
third-party
library.
Now
someone
else
also
bundles
that
third-party
library,
then
you've
got
a
conflicting
loader
again,
because
you
can't
anticipate
someone
else
providing
that
loader.
So
we
have
to
have
a
parallel
deduced,
the
loader
system,
in
order
to
rectify
that.
So,
if
even
if
it
was
scope
for
Sadie
3
or
something
someone
else
provided
III
and
different
logo,
same
loader
minikin
school
cause,
an
issue
can
webpack.
A
A
B
A
B
A
B
B
A
B
A
B
A
C
So
I
have
been
working
on
extracting
some
of
the
stuff,
that's
integrated
into
the
file
browser
out,
so
we
have
one
PR,
that's
already
merged,
that
creates
a
file
browser
factory
that
sort
of
makes
it
easier
to
create
them
because
you
don't
have
to
deal
with
passing
in
a
document
manager
or
services
manager
or
state
database
right.
You
just
can
call
the
fact
three
and
that
allows
you
to
more
easily
create
multiple
ones
and
move
the
state
restoration
of
all
browsers
out
of
the
extension
and
just
made
it
native
to
the
file
browser.
C
So
all
of
them
now
remember
the
last
folder
that
was
open.
Unless
you
explicitly
tell
you,
don't
want
it
to
do
that
when,
when
you're
in
satiating
it
via
the
factory
and
now
I'm
going
through
and
some
of
the
functionality
of
file
browser,
it
works
by
creating
dynamically
named
commands
on
the
fly,
but
instead
those
commands
really
ought
to
be
in
the
document
manager.
So
things
like
create
a
new
folder
create
a
new
notebook.
C
Those
things
should
really
be
part
of
the
document
manager
and
not
created
/
file,
browser
instance
and
so
pulled
out
of
the
file
browser
model,
so
that
that's
what
I'm
working
on
now
and
the
end
goal
is
so
that
there's
a
uniform
set
of
commands,
anyone
can
use
to
do
file,
manipulation
and
file.
Browsers
just
happen
to
be
one
of
the
clients
of
those
commands
yeah.
That's
it.
D
So
I
this
is
intended
to
get
some
heading
of
Jupiter
log,
which
is
that
I
opened
a
PR
earlier
this
week
on
on
a
fairly
simple
proposed
interface
for
model
database
backends.
So
the
idea
is
game
at
different
document
models
can
use
this
interface
and
then
those
that
that
database
can
be
real-time
enabled,
and
so
the
basic
structure
of
the
thing
is
just:
we've
got
factory
methods
for
our
various
observable
fraid
of
primitives,
so
each
each
back-end
would
need
to
be
able
to
implement
some
version
of
those
primitives
and
then
getters
for
those.
D
So
the
various
document
models
will
create
the
create
observables
that
they
need,
and
then
hopefully,
if
you
do
that,
then
everything
should
just
work
and
so
I
guess
I
could
use
some
eyes
on
this
on
this
proposal.
If
people
have
a
few
extra
cycles,
it's
a
big
PR
and
it
sort
of
does
three
things
so
ultimately,
we'll
probably
need
to
be
broken
up.
So
one
thing
that
it
does
is
implement
this.
This
model
database
and
other
thing
that
it
does.
D
D
A
A
C
A
Sure,
I
guess
one
other
bit
of
things
so
Brian
and
I
stood
submitted.
A
proposal
for
a
Jew
corolla
talk
two
strata
New
York,
because
because
it
was
submitted
minutes
before
the
deadline,
we
didn't
have
a
chance
to
get
headshots
of
cifra
Darien.
So
we
didn't
add
you
as
office
because
they
wouldn't
let
us
submit
it.
But
we'd
like
to
put
these
Darien
Chris
co-op
Azure
the
strata
New
York
talk.
I
will
try
to
append
you
as
authors.
If
we
get
accepted
so
and.