►
From YouTube: JupyterLab Weekly Dev Meeting, March 3, 2017
Description
Meeting of the JupyterLab development team, March 3, 2017.
Meeting Notes: https://jupyter.hackpad.com/JupyterLabNotebook-Weekly-Meetings-UUJ3gIQ3iBS
A
B
The
big
thing
that
happened
this
last
week
on
our
end
is
hi.
Widget
6.0
is
released,
so
we
got
all
the
packages
up
and
everything,
except
maybe
the
OS
X
packages.
We
should
check
on
that.
The
Condor
Forge
built
bot
was
delayed
for
OS
X,
but
it
seems
like
it's
relatively
smooth.
There's
a
few
questions.
We've
been
answering
compatibility,
questions
and
things
like
that,
but
it
seems
like
things
like
things
are
working
pretty
well.
C
B
Thanks
for
all
your
guys's
help
to,
hopefully
this
means
we
can
spend
a
little
more
time
on
Jupiter
web
and
but
our
first
order
of
business
will
be
now
porting
widgets
over
to
phosphor
1
packages.
So
right
now,
I
think
the
plugin
won't
work
with
the
current
release
of
jupiter
lab
because
it's
based
on
the
new
phosphor
packages,
so
the
merry-go-round
keeps
running
around,
will
get
it
updated.
A
Or
I
guess
I'm
up
next
spent
most
of
the
week
breaking
out.
What
is
the
Jupiter
common
folder
Google,
a
common
folder
in
two
separate
packages,
mirroring
what's
in
phosphor,
that
Coria
tools
and
the
dami
tools
and
then
worked
with
Chris
on
we
had
looked
at
this
tool
called
learner
that
Babel
uses
to
handle
multiple
packages
from
one
repo.
We
didn't
think
if
we
did
what
we
wanted,
because
they
were
using
it
in
a
way
that
there
is
a
single
version
of
everything.
A
So
if
you
make
a
release,
everything
gets
a
major
release
when
you
just
wanted
to
make
do
a
patch
release
of
one
of
your
constituent
packages,
but
learner
picked
up
this
independent
mode,
so
it
allows
you
to
version
everything
independently.
So
we
went
through
the
check
on
phosphor,
because
the
base
case
simpler
than
Jupiter
lab
is
setup
and
it
did
work.
Chris
has
got
a
great
workflow
going
I'll.
A
Let
him
talk
about
what
that
is,
so
I
think
we're
about
ready
to
make
the
flip
the
switch
on
jupiter
lab
and
break
it
into
separate
components,
and
it
should
be
a
lot
smoother
than
we
thought
it
was
going
to
be
because
lawrence
laura
is
providing
the
heavy
lifting
of
linking
everything
together,
making
sure
it's
coherent,
rather
than
us,
trying
to
do
anything
manually
so
yeah,
it's
it's
great,
I'm,
very
happy
with
it.
We
might
need
to
have
some
discussion
about
what
we're
packages
live.
A
I
think
it
might
make
sense
at
this
point
to
bring
services
in
since,
since
it
can
be
an
independent
packages,
but
still
in
the
repo
to
bring
services
into
Jupiter
lab,
because
it's
going
to
use
choreo
tools
from
Jupiter
lab
anyway,
and
it's
already
it's
already
released
in
the
Jupiter
lab
namespace,
so
I
think
it
makes
sense
to
absorb
the
repo
we
don't
we
don't
decide
that
right
now
and
then
I
suppose
Foreman
gets
the
ian.
We
can
talk
more
about
models,
real
models.
What
those
purposes
on
that
in.
A
Well,
if
we
did
that,
then
we'd
also
have
to
make
core
utils
part
of
Jupiter
and
I'm
a
duper
lab
I
guess
we
could
think
of
it
as
the
JavaScript
parts
of
Jupiter.
If
you
want
to
look
at
it
that
way,
so,
once
I
had
broken
out
Cori
tables,
it
made
more
sense
to
me
to
actually
absorb
it
into
the
Jupiter
repo,
but
still
a
lot
of
like
Colonel
gateway
to
install
it
separately.
So
core
utils
and
services
will
be
no
usable
repos,
but
anything
above
that
you
start
requiring
a
window
in
the
document.
C
B
I
was
looking
at
we're
looking
at
spitting
up
the
iPad
widgets
repo
too,
but
I
think
we'll
still
split
it
up
and
not
use
learner,
because
what
the
situation
we
have
is
multiple
Python
packages,
which
the
convention
in
Python
packages
is
one
package
per
repo,
much
more
so
than
than
the
JavaScript
packages.
So,
whilst.
A
Yeah
so
it'll
be
less
of
a
concern
when
one
when
they're
just
individual
packages
not
to
hold
the
whole
of
Jupiter
lab,
so
I
think
I
think
it's
fine,
as
is
right.
Now,
basically,
you'd
have
to
use
the
Python
Jupiter
lab,
but
actually
I
don't
know.
Let's
think
about
that
in
the
short
term.
Maybe
what
that
would
look
like
to
use
Jupiter
lab
in
the
Alpha
sense,
not
with
an
NPM
package,
but
it
would
be
a
moot
point
coming
here.
A
Maybe
maybe
just
we
would
just
press
forward
with
the
changeover,
because
it's
going
to
be
very
straight
forward
to
at
least
get
the
initial
versions
of
the
packages
out
and
then
we
can
and
then
we
can
start
thinking
more
about
how
to
clean
those
up
after
the
fact
yeah.
Maybe
maybe
it's
not
what
the
chases
I'm.
C
C
B
D
A
C
C
A
C
C
C
A
A
A
E
E
Well,
notebook
we
haven't
done
that
yet,
but
I'm
sure
we
may,
that
turned
out
to
be
a
bit
complicated.
But
now
that
works
fine,
which
is
good.
It
means
you
can
just
anywhere.
You
have
an
editor.
You
could
have
completion
functionality
without
worrying
about
its
context
really,
and
the
other
thing
is
I
deleted,
an
underscore
that
turned
out
to
be
important.
So
we're
gonna
have
to
do
a
back.
Ported
patch
release.
C
C
So
this
this
issue
that
that's
listed
here,
some
of
the
common
actions
would
be
like
run,
sell,
clear,
sell,
output
and
then
I'm
not
sure
all
these
should
be
personal
actions
and
I've
been
working
to
work
on
some
of
the
UI
questions
about
how
to
expose
that
I
don't
want
to
adorn
cells
with
a
lot
of
UI.
It
needs
to
be
extremely
subtle
and
minimal
in
some
of
it
may
even
be
a
right-click
context
menu
and
then
so.
My
question
is
from
an
implementation
standpoint.
C
The
this
type
is
why
I,
don't
think
should
be.
In
the
cell
view,
it's
more
notebook
level
stuff.
Now
some
of
the
actions
would
be
per
cell,
some
of
like
clear
output.
Others
would
be
at
the
notebook
level.
For
example,
if
we
wanted
a
button
bit
that
just
inserts
a
code
cell
or
markdown
cell
below
that
cell
or
delete
so
well
delete,
so
I
guess
notebook
level,
action
or
cell
level
action
might.
A
C
It's
probably
going
to
be
some.
The
the
appearance
of
duis
will
probably
be
triggered
on
hover
like
one
example
that
I
played
with
just
visually
would
be.
If
you
hover
over
the
prompt,
the
input
prompt,
it
turns
to
a
play
button,
and
you
can
just
click
that
to
run
this
off
like
that.
That
would
be
one
example
of
the
type
of
thing
I'm
exploring
or.
B
C
So
there
I
mean
another
would
be
like
the
material
design
has
a
nice
more
icon,
it's
like
basically
triple
box,
either
horizontal
or
vertical
and,
like
a
user
can
hover
over
that
click
it
and
then
we
could
expose
some
small
micro
panel
of
actions.
Well,
there's
you
know
and
I
I'm
trying
here
to
ask
more
the
implementation
of
sort
of.
If
we
want
to
start
to
play
with
these
types
of
things,
is
there
a?
Is
there
an
API
or
what
does
that
look
like?
Where
would
we
start
to
plug
that
in
at
the
code
level.
A
B
E
B
Yeah
yeah,
if
we
have
an
abstracted
out
faster
widget,
say
that
does
the
attached
to
Dom
at
the
root
node,
you
know
floats
above
things
at
a
given
position,
etc.
That
would
be
that
would
be
really
handy
for
us
and
I
play
widgets
as
well.
You
know
it
handles
those
coral
events
and
stuff
like
that.
I
yeah,
yeah.
A
E
Actually,
hover
box
is
actually
just
a
function
called
set
geometry,
really,
the
scroll
events.
What
you
do
is
you,
as
the
owner
of
the
thing,
that's
gonna
hover.
All
you
do.
Is
you
attach
document
wide
scroll
listener,
not
on
a
specific
moment
on
the
whole
document,
and
then
you
make
your
scroll
listener.
E
B
E
E
But
if
I
have
that
Dom
you
till
I
could
easily
build
a
phosphor
widget
on
top
of
that
Dom
util
that
that
internally
uses
it
this
way,
and
then
you
could
use
that
phosphor,
which
it
to
do
whatever
you
like
I,
still
think,
there's
a
benefit
to
keeping
that
function
Dom
only,
but
if
you
want
to
foster,
would
build
one
on
top
of
it.
Yeah
yeah
I
mean
those.
A
E
Know
what
would
happen
the
problem
and
just
put
because
I
already
have
a
namespace
called
hover
box
with
that
function
in
it,
I
would
create
a
class
also
called
hover
box,
because
that
sounds
like
the
name
of
a
widget
anyway,
and
you
could
instantiate
either
a
new
hover
box
or
call
hover
box.
That
said
geometry,
would
you
like
I'm,
sorry
Chris?
What
did
you
just
say?
I.
G
B
B
You
got
to
be
careful
with
having
something
relative
inside
the
Dom
nodes,
because
if
you
have
a
parent
that
is
overflow:hidden
it'll
get
cut
off,
so
that's
I
mean
that's
the
issue.
We're
we're
we're
finding
right
now,
for
example,
in
beak,
you
fought
with
our
hover.
Things
is
that
if
we're
inside
of
a
tablet
inside
of
ipi
widgets
with
overflow:hidden,
then
it
gets
cut
off
because
it's
cut
it
clipped
to
the
boundary
of
the
container.
So
if
you
want
something
that
really
is
hovering
over
everything,
then
then
you'll
want
something.
B
C
B
B
C
C
Others
of
it
might
appear
anytime,
you
hover
over
a
cell,
even
if
it's
not
selected
and
so
I
and
I.
Think
part
of
my
question
is
like
I
I
think
this
UI
is
going
to
have
to
be
very
tightly
integrated
with
the
Dom
structure
of
the
cell,
like
it
may
even
involve
changing
the
layout
of
the
cell
itself
in
some
way.
So,
for
example,
like
you
could
imagine
there
being
in
the
far
right
of
a
cell
dot-dot-dot
indicate
more
when
you
click
it.
C
In
it,
and
so
or
you
know
like
replacing
the
prompt
with
a
play
button
when
you
hover
over
it,
like
that's
a
pretty
that's,
not
some
other
thing
that
you're
putting
on
top
of
the
cell
you're
literally
mutating
the
Dom
of
the
cell
itself,
and
you
know,
maybe
we
need
a
subclass
of
the
cell
view
that
has
this
extra
UI.
That
may
be
the
answer.
A
A
E
That's
right,
you
can
there's
this
concept.
I
kind
of
wish.
I
hadn't
used
this
word,
but
it
seems
like
the
right
word
of
namespaces.
You
give
the
state
database
namespace
and
then
everything
that
it
stores
is
stored
with
that
as
a
prefix.
But
it's
totally
transparent
to
you.
You
just
create
a
new
state
database
with
namespace
equals
clue.
And
then,
when
you
add
keys,
you
don't
give
the
flu
you
just
give
the
bar
or
whatever,
and
when
you
get
bar
it
actually
looks
inside
state
database
for
food
out
bar,
and
so
that's
already.
E
In
the
URL
yeah,
you
can
have
a
URL,
fragment,
I.
Guess
here's
the
thing
about
URLs.
We
should
be
very
careful
about
what
we
put
into
URLs.
They
have
a
way
of
lasting
forever.
If
we
make
a
mistake
today
that
that
might
have
consequences.
So
certainly
something
belongs
in
the
URL.
Some
amount
of
our
state
has
to
go
there
because
we
want
to
share
things
with
people.
I
just
think
it
requires
real
discipline
because
you
could
easily
cram
a
lot
of
stuff
in
there
because
it'll
solve
a
problem
really
quickly.
C
E
You
could
store
it
in
a
state
database
itself.
You
can
store
it
in
the
main
application
level
database.
You
could
store
it
in
whatever
setting
system
and
the
building
or
you
can
store
it
in
a
cookie
or
you
could
not.
You
could
have
all
the
different
ones
available
to
you
in
some
sort
of
drop-down
UI
that
is
read
from
the
state
database,
but
keep
in
mind
state
database
every
single
time.
There's
a
version
change
will
delete
everything.
E
E
D
C
E
E
B
Which
yeah,
you
probably
remember
these
day
me
that
the
jlabbe
URL
should
have
a
session
ID
and
it
should.
It
should
be
a
refreshing
thing
anyway.
Those
are
just
preliminary
about
it
all.
Obviously,
it
would
be
far
beyond
that
with
darienne's
actual
work
on
it,
but
I
put
it
in
a
comment
and
CCG
ilanic.
E
G
Updating
the
build
system
to
use.
Lerna
Steve
had
mentioned
that
a
bit
earlier
learner
is
basically
a
tool
that
assists
in
managing
mono
repos,
like
what
we
have
where
you
have.
One
can
help
repo
that
manages
a
bunch
of
it
and
then
NPM
packages,
and
so
it
really
drastically
simplified
our
build
release
cycles.
G
So
what
we
were
really
waiting
for,
but
it
didn't
have
the
first
time
we
looked
at
it,
was
that
independent
mode
which
allows
us
to
independently
version
the
packages
within
the
repo
rather
than
having
one
single
version
for
every
package
of
their
repo.
So
that
took
a
little
bit
of
reorganizing
the
repo
itself.
No
api's
change
I
just
needed
to
move
kind
of
the
structure
of
things
like
where
the
tests
lived
and
what
the
names
of
the
folders
were
updated.
G
Some
of
the
scripts
in
the
package,
JSON
files,
but
I-
think
I
sent
along
a
pattern
that
that
appears
to
work
pretty
good.
So
that's
it
master.
Now,
if
people
kind
of
want
to
see
what
things
look
like,
also
spent
a
little
bit
of
time,
implementing
some
very
quick
community
feature
requests
some
stuff
around
JSON
utils
there,
the
JSON
utilities
and
a
couple
other
things.
They
reminder,
though,
compared
to
the
the
repo
maintenance
I've
got
another
one.
That
kind
of
popped
up
today
from
Anton
was
asking
an
additional
feature
to
signaling.
G
That
would
allow
you
to
disconnect
basically
break
all
the
connections
between
a
given
sender
and
receiver.
So
there's
an
issue
for
that.
If
somebody
wants
to
chime
in
about
what
that
function
should
be
named,
it's
a
reasonable
request,
so
I'll
add
it
and
it's
easy
to
add
its
need
to
figure
out
what
that
come
should
be
named.
G
So
beyond
that
once
I
finish
this,
this
feature
request
I'm,
going
to
attempt
to
use
Leona
to
push
out
new
releases
of
everything,
and
so
we'll
get
that
process
going
so
I.
Imagine
that
everything
whether
it
eats
a
bump
or
not,
is
going
to
get
a
bump
just
to
get
it
synchronized
with
learner
and
they
get
tags.
So
we
can
move
forward
with
those
synchronize
releases,
so
I'll
try
to
get
that
get
that
out
today
and
then
other
than
that
I'd
say
we're
ready
for
a
one
dot,
oh
release.
G
Unless
you
want
me
to
wait,
so
you
can
make
sure
that
the
iPod
widget
supports
goes
well.
We
you
know
made
it
catch
bugs
there
or
is
in
what
we
had
in
some
of
these
dock
panel
features.
We
were
talking
about
before
the
window
to
release
like
maximization
and
tar
out,
and
they
were
store
layout,
so
I'm,
basically
ready
to
kind
of
wand.
I
know
whenever
you
guys
say
you're
ready,
and
why
is
what
a
basically
we're
in
that.
B
I
would
say:
I'm
gonna
return
to
getting
my
parrots.
You
know
forking
it
so
that
we
have
a
stable
version
at
six
and
then
moving
the
dead
version
over
to
the
new,
the
new
phosphor
yeah.
If
you
could,
if
you
could
hold
off
while
I
at
least
do
that
and
we'll
see,
if
there's
any
other
bugs
that
I
encounter
doing,
that,
that'd
be
great
I'm
in
favor
of
cutting
a
release
before
doing
the
features.
B
Just
just
because
we've
been
waiting
for
this
for
a
long
time
and
we
can
always
cut
a
2.0
if
we
need
to
do
you
see
the
features
that
the
usually
we're
talking
about
the
maximize
and
the
tear
out
backwards,
incompatible
yeah,
so
Steve.
What
are
your
thoughts?
I'd
rather
see
release,
then
to
hold
off
more
but
yeah.
C
C
C
If
not
we'll
just
try
on
that
note,
we
have
been
talking
about
what
a
good
abstraction
is
for
for
having
a
database
for
live
models
and
sort
of
there's
some
contention
about
what
the
what
appropriate
data
structures
can
be.
Don'ts
on
this
model,
anything
ranging
from
only
Jason
double
values,
up
to
Jason
values,
plus
editable
strings
and
having
successfully
Richard
Richard
models
which
carry
with
location,
deep
problems
but
may
be
necessary
to
provide
the
kind
of
granularity
we
need.
C
So
so
what
one
thing
were
were
we're
going
to
attempt
today
and
Steve
you
might
want
to
join
in
is
to
have
all
of
us
Cameron
myself
and
Ian
install
the
current
branch
that
has
real-time
collaboration
and
attempt
to
use
it
today
to
keep
our
notes
in
a
markdown
document
to
sort
of
both
start
to
play
with
the
user.
Experience
has
some
of
the
performance
issues
and
so
like
we
may
end
up
keeping
our
notes
using
Google
Drive
and
a
collaborative
markdown
document
today.
C
B
The
true
Peter
Kahn's
mission
deadlines
coming
up
and
sci-fi
is
coming
up
after
that
submission
deadline.
Jupiter
con,
we
were
gonna,
submit
a
tutorial
on
with
the
widget
ecosystem
I,
but
I
think
we
should
definitely
have
a
talk
on
Jupiter
lab
and,
if
not
a
tutorial
on,
writing
Jupiter
lab
widgets,
which
may
be
very
interesting.
Maybe
it
may
not
be
broad
enough
to
actually
get
accepted,
but
but
yeah.
C
I
can
give
a
little
update.
I
had
a
brief
conversation
with
Fernando
like
9
or
10
o'clock
last
night
about
the
the
idea
of
programmatic
talks,
namely
sort
of
tops
that
are
essentially
invited,
but
they're
not
sort
of
traditional
O'reilly
style
keynotes
and
the
idea
would
be
to
have
our
core
devs.
Give
talk,
give
talks
on
the
core
architectural
pieces
of
the
project
or
like
the
message,
spec
notebook
documents
and
I'm
guessing
a
sort
of
an
overview
of
Jupiter
lab
would
be
one
of
those
that
that
still
need
to
be
worked
out.
C
I
think
the
format
of
those
if
I,
remember
for
nonde
staying
correctly
would
probably
be
40
minutes,
but
I
I
definitely
think
having
a
submit
a
to
toriel
on
building
easily
building
extensions
for
jupiter,
lab
or
jupiter
lab
development
would
be
a
really
good
idea.
In
addition
to
whatever
that
sort
of
programmatic
talk
look
like
so.
D
C
I
think
that's
I
mean
I
think
we
should
definitely
have
a
talk
on
jupiter
lab
and
if,
if
it
turns
out
it's
not
one
of
those
programmatic
talks-
and
I
don't
I'm
not
worried
about
us
submitting
or
doing
like
getting
something
in
through
another
route,
I
do
think
those
submit
submitting
a
jupiter
lab
development.
Extension.
Development
tutorial
is
a
good
idea,
and
so
we
should
probably
think
online
about
that
and
figure
out
sort
of
who
would
want
to
participate
and
how
we
approach
that
no.
B
And
same
sort
of
kind
of
I,
don't
think
we
need
to
do
a
tutorial
for
sign
five,
but
we
should
have
another
Jupiter
lab
talk,
probably
a
sign
pie.
So,
let's
sync
offline
about
who's
planning
on
going
to
scifi
and/or,
the
various
PI
data,
PI
data,
Seattle
I,
think
is
the
week
before
site
I.
You
know
just
to
coordinate
whatever
open
up
clocks
and
Darien
and
I
submitted
a
talk
to
PI
data
London
for
to
Fodor
lab.
C
C
F
So
let
me
just
dropping
the
events
calendar
into
the
chat,
so
these
are
all
the
upcoming
talks,
I'm
sorry
upcoming
conferences
that
we
attend
in
some
shape
or
form
so
the
ones
that,
coupled
with
the
tosses
Jupiter
events
I've
added.
So
if
folks
for
the
folks
who
are
attending
these
conferences,
if
you
could
assign
yourself
to
not
card
so
we
know
who's
going,
that's
the
issue.