►
From YouTube: JupyterLab Dev Meeting, January 6, 2017
Description
Meeting of the JupyterLab Development Team, January 6, 2017
Meeting Notes: https://jupyter.hackpad.com/JupyterLabNotebook-Weekly-Meetings-UUJ3gIQ3iBS
A
Hi,
everyone
welcome
to
the
Jupiter
lab
team
beating
for
Friday
January.
Sixth,
and
this
is
our
first
team
meeting
for
Jupiter
lab
in
2017.
All
of
us
have
taken
some
time
off
and
we're
more
or
less
all
plugged
back
in.
At
this
point,
we've
got
some
people
might
maybe
late
they're
over
it
visiting
Bloomberg
and
working
there
for
the
day.
But
if
you
have
some
fun,
so
why
don't?
We
begin
and
I
guess
I'm
first
in
the
hot
pad,
haven't
done
much
mostly
doing
issue
for
Yas
for
Jupiter
lab
and
lonely
going
through
the
issues.
A
At
this
point,
I'm
trying
to
go
component
by
component
and
in
particular
I'm
going
in
order
of
number
of
issues
open
for
for
the
component
and
in
the
half
have
there
I've
got
an
issue
that
I've
opened
where
I'm
posting
visualizations
of
the
number
of
issues.
We
have
sort
of
help
us
get
a
big
picture
and
they're
basically
bar
charts
of
the
count
of
the
number
of
issues
grouped
by
the
milestone
and
then
also
grouped
by
the
component,
and
so
that
right
now.
A
So
right
now
in
the
beta
milestone,
the
the
biggest
component
we
have
is
the
notebook
in
terms
of
number
of
issues
open.
Then
the
application,
then
the
file
browser
the
build
system
and
then
a
small
number
of
issues
for,
for
other
components
in
the
overall
counts:
the
notebook,
the
application,
the
file
browser,
the
editor
console
terminal,
those
types
of
things,
but
tend
to
be
the
highest
count,
but
I'll.
Let
everyone
look
at
the
actual
issue.
A
Many
thanks
to
everyone,
who's
been
helping
with
issue
triage.
Are
our
issues
are
actually
really
really
good
condition
on
Jupiter
lab
we've
got
two
hundred
and
something
issues
open
right
now
to
14
and
the
main
thing
is
they're
very
well
organized
I,
don't
think,
there's
many
that
are
sitting
there
dying,
not
getting
attention,
and
so
that's
really.
A
team
effort
and
I
just
wanted
to
say
thanks
to
everyone
for
working
so
hard
on
that
it
makes
a
huge
difference
in
terms
of
planning
and
tracking.
A
All
this
I
am
finding
as
I
go
through
components
for
the
beta
release,
because
our
issues
are
already
so
well
triaged.
We
don't
have
many
that
are
that
we
can
close
because
they're
either
done
or
no
longer
relevant,
and
what
that
means
is
that
as
I
do
this
pass?
The
number
of
beta
issues
is
likely
to
go
up
slightly
as
I
sort
of
try
to
figure
out
what
answer
the
question?
What
do
we
need
to
do
to
get
each
component
ready
for
the
release?
A
The
other
thing
that
I've
done
is
I'm
teaching
next
week
and
I
was
able
to
very
easily
get
Jupiter
lab
installed
with
Jupiter
hub.
That
was
basically
Conda
installed
Jupiter
lab,
and
that
was
done.
The
only
thing
that
that
I
don't
have
yet,
and
for
me
it's
not.
A
big
deal
is
the
extra
buttons
that
Jupiter
hub
puts
in
the
notebook
for
logging
out
and
the
control
panel
again
Alan's
working
on
a
PR,
and
you
know
Jupiter
hub
repo.
A
That
adds
those,
but
users
can
always
go
to
the
classic
notebook
page
to
access
those
buttons
for
now.
So
I
am
going
to
offer
as
an
alternative
to
the
students.
They'll
probably
end
up
working
both
in
the
classic
notebook
and
into
her
labs
throughout
the
quarter.
But
I
was
very
very
pleased
that
it
just
magic.
A
We
weren't
the
only
thing
that
I'm
a
little
bit
puzzled
about
that
at
least
wanted
to
bring
up
is
I
had
opened
up
an
issue
about
not
being
on
access
lab
when
I
had
started
the
notebook
server
using
Jupiter,
/
space,
notebook
and
Steve.
You
had
commented
that
you
think
Jupiter
lab
has
to
be
enabled
in
the
user
config.
Rather
the
system
and
the
sis
prefix
config,
but
I'm,
not
finding
that
on
my
Jupiter
hub
deployment,
I've
enabled
everything
in
sis
prefix
and
it
all
works.
A
B
A
Okay,
I
I
was
a
little
surprised.
It
just
worked
with
sis
pre-fit
on
Jupiter
hub,
but
there's
nothing
special
about
Jupiter
well,
I
suppose
I
mean
when
you
run
the
notebook
on
in
Jupiter
hub.
There
is
a
separate
top
level
at
single
user
notebook
app
that
Jupiter
hub
has
subclasses
the
classic
notebook,
app
and
I.
Suppose
there
could
be
some
very
subtle,
config
difference,
handling
that
that's
there
at
least
wanted
to
bring
it
up
with
folks.
All
all
I
mean
the
good
news
is
Jupiter
hub
is
working
great
with
it.
A
C
Hi
so
I've
picked
up
where
I
left
off
with
stay
in
session
restoration
I
had
to
do
some
rethinking
about
where
the
different
responsibilities
lie
in
the
life
cycle
of
from
the
moment,
the
page
starts
loading
to
the
moment
the
layout
actually
gets
applied.
So
the
process
now
is
that
there
is
a
state
database
plugin.
There
is
a
plugin
called
an
instance
restorer,
which
is
dependent
on
the
state
database
and
then
all
other
plugins
that
put
widgets
on
the
page
are
dependent
on
the
instance
restore
they.
C
They
ask
the
wooden
instance
restore
to
restore
all
of
the
widgets.
They
had
been
tracking
before
using
their
instance
tracker
and
as
each
one
of
those
finishes.
The
instance
restore
keeps
track
when
every
single
plugin
that
had
asked
it
to
restore
things
has
finished.
It
then
fires
off
a
restored,
a
promise.
C
C
Dehydrated
status
state
of
the
main
area
as
well
right
now
we're
only
really
dehydrating
the
left
on
the
right
area.
That's
the
conceptual
model,
and
it's
almost
there,
I'm
hoping
actually
before
I
knock
off
tonight.
To
have
it
done,
but
I
would
suggest
everyone
once
it's
out
there
sort
of
pay
attention
to
idiosyncrasies,
because
there's
there's
a
really
well
defined
life
cycle,
but
I
think
it's
easy
to
get
things
wrong.
So
I
want
to
I
want
to
make
sure
it's
working
right
and
it's
still,
the
exact
same
burden
exists
on
plug-in
developers.
C
As
last
time
we
talked
about
it,
there's
nothing
new
that
a
plug-in
developer
needs
to
do
all
they
have
to
do
to
make
this
work
is
to
use
the
instance
tracker.
However,
the
one
thing
that
we
aren't
doing
that
we'll
start
doing
is
every
single
widget
on
the
page
that
attaches
itself
to
the
app
shell
should
use
an
instance
tracker
that
includes
the
main
menu
that
includes
a
bottom
area.
C
If
we
start
using
it
and
think
that
you
that
says,
she'll
add
to
foo
area
should
be
using
an
instance
tracker,
which
is
fine,
because
for
now
we
control
all
those
and
then
once
people
start
building
and
they'll
have
our
models
to
work
off
of,
and
then
everything
else
should
show
to
magically
put
itself
together.
So
that's
what
that
is
and
I
think
once
once
this
PR
goes
in
I
think
we
have
beta
level
support
for
state
restoration,
acceptable
enough
that
it
won't
annoy
beta
users
like
they.
C
C
Yeah,
that's
a
good
point.
I
think
we
have
some
plugins
that
are
really
short
like.
If
you
look
at
some
of
our
plugin
TS
files,
there
may
be,
you
know,
10
15
lines
and
I
think
if
that
the
boilerplate
to
creating
a
plug-in
that
puts
a
widget
somewhere
on
the
page
that
is
restorable
is
actually
fairly
small,
so
that
that
might
be.
That
might
be
nice.
It
might
even
be
the
case
not
a
hundred
percent
sure,
but
it
might
even
be
the
case
that
we
could
have
something
like
a.
C
Those
are
the
those
are
the
absolute
bare
necessities
to
be
able
to
refresh
these
things
and
to
be
able
to
serialize
them.
So
you
can
imagine
augmenting
our
ad
to
main
area
at
the
right
area
at
the
left
area,
to
take
a
second
argument
that
basically
gives
those
things
and
then
just
abstracting
away
the
need
for
users
to
use
the
instance
tracker
at
all.
Maybe
I'm
not
sure
it's
worth
talking
about,
and
it
could
simplify
things,
but
it
could
also
be
it
even
though
it
might
be
fewer
lines.
C
B
Alright,
so
this
week,
I
took
took
on
the
revisit
of
the
abstract
editor,
while
I
was
running
tests
for
the
code
mirror
implementation
I
realized
that
we
were
misusing
the
code
mirror
doc
in
that
it
is
intricately
tied
to
a
code,
mirror
editor
in
that
it
can't
be
used
in
a
multi-user
situation.
So
real-time
collaboration
wouldn't
work
where
you
have
it
that
doc
as
a
model,
because
it
has
its
own
concept
of
cursors
and
he
can't
have
foreign
cursors
so
that
led
to
a
rethinking
and
actually
cleaned
up
a
lot
of
logic.
B
Where
now
the
model
is
just
a
description
of
code
itself,
it's
just
the
the
the
source
code
am
on
type
and
then
this
selections
object
that
describes
the
cursors
and
selections
from
different
users
are
using
they're
looking
into
that
model,
and
what
that?
What
that
turned
out
to
look
like
is
that
a
default
document
that's
created.
The
document
model
is
a
code
editor
model,
so
that
can
be
directly
to
the
editor
widget
and
it
just
uses
that
to
create
a
code.
Editor
instance
to
go
with
that,
so
the
code
mirror
is
using
that
model.
B
I
think
this
will
make
the
real-time
collaboration
easier
on
Ian
and
it
turns
out.
We
can
completely
get
rid
of
this
thing
now
that
we
had
called
it
the
coats,
the
cell
editor
and
that
there's
nothing
special
about
it.
That's
in
the
next
pass
will
be
to
remove
that
thing,
which
is
now
just
three
signals
on
top
of
the
standard
in
line
editor,
that
our
completion
requested,
this
text
change,
state,
amit
and
an
edge
requested,
so
those
those
can
move
on
to
the
abstract.
B
B
The
process
right
now
breaking
phosphor
into
logical
sub
modules
so
that
you
can
import
from
a
given
sub-module
and
those
are
independently
Persians
and
the
reason
to
do
this
versus
what
we
have
been
doing
was
that
vehicle.
If
we
make
the
level
of
import
a
module
than
we
were
saying
that
that
that
module
is
simple,
we're
using
that
for
our
son
birdie
duping
and
turn
out
that
that
was
not
a
valid
way
of
doing
it.
B
Because
you
can,
you
can
logically
de
doop
a
package
at
the
module
level
has
to
be
done
at
the
package
level,
so
we're
parsing
out
what
what
is
a
package
to
sub
package
and
doing
at
the
phosphor
level,
and
then
that
that
PR
or
that
is,
you
is
tracking.
What
we
need
to
do
to
make
it
clean
at
the
Jupiter
lab
level
and
that
will
make
it
easier
for
lab
dependencies
to
be
properly
d
duped
at
the
client
level,
because
we're
d
duping
full
packages
that
are
this
package
is
atomic
versus
trying
to
say.
B
A
B
It's
actually
the
intent
is
that
each
folder
each
sub
folder
is
is
a
is
a
sub
package.
So
it's
only
fall
unique
document
registry.
You
would
import
off
in
the
registry
and
that's
the
thing
that
you
would
have
to
be
doop
and
the
client
and
those
each
of
those
will
have
independent
versions.
Yes,
so
chris
has
got
a
great
solution
for
automating
that
as
much
as
we
can,
but
still
it's
still
a
coherent,
sensible
thing.
B
It
requires
some
discipline
on
the
part
of
the
the
package
author,
but
it's
already
a
part
of
is
as
long
as
you're
following
stumberg.
The
the
process
works
so
well.
We'll
have
more
we'll
talk
more
about
that
once
it's
implemented
in
phosphor
or
Chris.
Chris
can
talk
more
about
that
now.
If
you
want
okay.
D
It's
still
altima
tlie
getting
flushed
out,
okay
on
my
end,
but
so
far
it
seems
to
be
saying
it's
definitely
easy
to
develop
so
like
if
I
need
to
make
a
change
in
one
package
that
I
can
consume
from
another
package.
It's
nice
that
I
can
make
the
change
and
test
it
all
locally
and
get
it
all
running.
D
It
makes
for
everything
compiles
insistently
before
publishing
any
actual
packages,
which
is
the
problem
where
you
have
a
different
repo
per
package,
because
you
have
to
actually
make
the
change
and
publish
and
then
install
it
in
your
new
ended
package.
And
oh,
it
wasn't
actually
the
change
I
wanted.
So
now.
Let
me
make
it
change
yet.
So
we
eliminate
that
so
that
we
don't
have
that
pain
anymore.
D
But
we
do
have
a
little
bit
of
you
have
to
be
a
bit
disciplined
when
releasing
packages,
because,
basically,
what
ultimately,
what
happens
is
that
typescript
has
a
global
view
of
the
repository?
And
so
when
you're
compiling
packages
typescript
thinks
that
whatever
is
currently
in
your
repo
meeting
trunk?
It
is
what
it
is
like
that's
its
view
of
the
world.
D
So
you
can
run
into
a
situation
where
you're,
not
where,
if
you're
not
disciplined,
that
you
could
fail
to
publish
a
dependency
package,
and
so
the
consumers
of
your
published
package
have
a
different
view
of
the
world
that
the
typescript
compiler
had
when
it
compiled
everything
and
said
everything
was
okay.
And
so
you
have
to
be
a
bit
disciplined
in
that
respect
and
make
sure
that
your
publishing
your
view
of
the
world
like
at
the
time
the
compiler
runs.
But
it's
the
same.
B
A
B
B
C
C
B
C
C
B
Are
the
same
way,
Babel
mono
repo
packages
are
so
it's
very
much
mirrors
that
structure,
except
the
only
difference,
is
we're
not
using
the
single
version
across
the
different
sub
modules,
because
that
forces,
if
you
make
a
patch
release
the
one
that
they
all
have
to
be
patch
released,
so
we're
trying
out
this
cember
managed
compromise
and
see
how
well
it
works.
Sometimes
we
can
be
disciplined
about
it
if.
D
The
thing
I
don't
like
about
that
is
that,
like
let's
say
like
collections,
the
phosphor
collections
package
gets
stable
and
we're
not
anything
to
it.
But
you
know,
every
change
is
that
I
maked
phosphor
UI,
which
is
going
to
be
often
as
we
had
more
features.
Then,
when
I
release
now,
all
those
other
stable
packages
get
bumped,
because
I'm
running
is
a
version,
and
I
don't
want
to
do
that
and
so
because
it's
silly
to
have
like
you
know,
20
different
versions
of
the
same
identical
code
of
a
package,
so
no
to
leave
yeah.
B
B
F
B
Last
piece
I
had
was
right
now,
internally,
the
archer
are
our
tests
in
jupiter
lab
are
not
compatible
with
notebook.
Four
point
three
point,
one
which
led
me
to
look
at
alternative
testing
solutions.
I
tried
out
selenium
and
the
great
thing
about
selenium
is
that
you
could
be
on
the
same
origin,
but
in
practice
and
I
have
a
right
up.
I
made
a
test
repository
as
a
proof
of
concept.
B
So
it's
a
bit
chromedriver
a
binary
that
you'd
have
to
install.
It
was
just
an
untenable
approach
to
force
every
user
install
to
do
this
and
it
limited
the
browsers
we
could
test
against
so
and
then
I
do
have
a
solution.
I
believe
to
getting
us
compatible
with
four
point
three
point:
one
without
having
to
resort
to
turning
off
all
the
security
flags,
I'm
almost
there
with
that
arm,
but
the
this,
the
selenium,
is
still
a
great
tool
for
if
we
want
to
say
make
sure
that
the
g
page
locally
on
trends,
we
could.
B
We
can
spin
to
the
page
and
check
some
content
in
the
cage,
make
sure
it's
their
kind
of
thing
and
the
same
with
our
examples,
we
can
make
sure
the
example
actually
properly
loads.
So
as
a
travis
test,
I
still
think
it's
a
viable
thing,
but
in
terms
of
using
it
for
our
unit
tests,
I
don't
see
us
that
being
a
terrible
solution.
A
B
C
Brian
did
the
basic
problem
is
that
the
moco
tests,
the
way
they
run?
Is
they
pass
through
all
your
requests
through
a
mocha
server
to
the
actual
Jupiter
server?
And
so,
if
we
have
to
get
a
token
from
the
server
and
give
it
back
and
get
a
cookie
in
order
to
avoid
a
cross-site
forgery,
vulnerability
that
breaks
with
the
new
version,
because
we
effectively
have
a
proxy
between
us
and
the
actual
API
server
that
wasn't
an
issue
before,
because
we
didn't
have
that
token
requirement
and
that
that
vulnerability
patch.
C
So
between
four
point,
three
point:
oh
and
four
point
three
point:
one
our
tests
broke
because
we
had.
We
always
have
this
server
in
between.
So
one
of
the
things
that
Steve
is
done,
that's
actually
kind
of
Awesome
is
instead
of
running.
Instead
of
having
this
proxy
server
running,
our
tests
he's
got
basically
a
extended
a
server
Jupiter
server
running
our
tests,
so
he's
got
a
separate.
That's
the
test
app
and
it's
a
lot
closer
to
the
reality
of
an
environment.
C
B
D
B
So
we're
there,
we
had
it's
just
if
we
get
stuck
with
how
do
we
how
to
get
the
data
back
out
of
the
page,
although
yeah,
what
we?
What
we
could
do
is
have
an
attribute
of
the
page
that
is
in
the
console
log
and
then
make
the
mocha
reporter
right
to
that
location,
but
we
would
lose
the
actual
the
console
logs
as
a
result
of
running
the
tests
themselves.
B
A
B
Seems
like
a
really
yeah
I
asked
min
and
there
is
a
flag
for
it,
so
yeah
we
can
completely
disable
these
these
checks,
it's
just
a
matter
of.
Do
we
want
to
do
that
going
forward
made
adjusted
the
tests
right
right,
but
it
also
signifies
that
you
purchase
services
is
not
compatible
with
notebook
focal
integral
one.
If
you
wanted
to
use
it
f4
in
something
like
supernal
gateway,
for
instance,
so
that
is
broken
right
now,
because
we're
not
compatible
okay.
B
A
B
A
A
A
E
F
Yeah
I
was
just
going
to
mention:
we've
been
working
a
lot
on
that
pie.
Widgets,
though
there's
been
a
lot
of
time
off
to
know
someone's
been
working
a
lot
more
than
I
have
over
the
last
couple
of
weeks.
I
did
backport
the
output
widget
to
4.3
point
ones,
so
iPad
widgets
6
beta
was
relying
on
a
dev
master
for
notebook,
but
now
it
relies
on
4.3
port
1,
plus
a
PR
or
death
master
which
which,
depending
on
the
really
schedule
for
notebook,
may
mean
we
want
to
make
4.4
release
with
these
changes.
F
E
E
We
have
a
few
new
issues
since
then,
and
it's
like
funny
now,
I'd
like
to
make
public
beta,
but
not
really
public,
just
public
to
people
who
actually
are
authored,
like
complex
third
party
widget
libraries
so
like
to
reach
out
to
be
autodesk
folks
and
the
ng
of
you
folks,
so
that
then
try
it
out
and
paris
if
there
is
any
like
huge
brokers
for
them,
because
between
five
and
six
we
we
just
change
the
programming
language
at
I/o,
justice,
freedom
and
maybe
I
mean
they
find
some
things
right.
Yeah.
A
So
I
think
you're
probably
release
I
by
widget
before
notebook
50
by
a
bit
part
I
mean
I
think
I
looked
last
night.
There's
150
issues
open
on
notebook,
50
I,
don't
think
all
those
have
to
be
addressed,
but
we
at
least
have
to
go
through
them.
The
biggest
issue
is
that
we're
reverting
all
the
web
pack
stuff
and
there's
a
lot
of
third-party
extensions
and
B
extensions
that
are
broken
on
master
and
so
I
think
we're
were
ways
off
still
and
so
I
I'm
guessing.
E
Yeah,
we're
gonna,
keep
it
your
lab.
We
need
to
discuss
the
status
of
the
lab
extension
that
we
we
have
in
I
by
widgets,
so
I
guess
that
it's
going
to
be
tearin
up
to
do
beta,
because
we
depend
on
Jupiter
lab
zero
point.
Whatever
is
released
when
we
do
release
it,
and
if
the
Jupiter
lab
team
makes
a
new
version,
that's
backward
compatible
at
the
next
two
weeks,
then
that
package
is
not
going
to
be
usable
anymore.
Can
you.
A
E
Yeah
we've
been
publishing
NPM
and
Python
package
at
different
times
already,
when
we
made
we're
making
like
minor
patch
versions
of,
for
example,
just
the
extension
and
not
the
widgets.
At
the
moment,
we
have
widgets
and
extension
Jupiter,
app,
widgets
and
I
by
widgets
as
Python
packages,
and
we
have
four
NPM
packages
in
the
same
repo.
So
that'd
be
great.
If
we
could
have
the
same
big
split
as
the
notebook.
No,
no
that
we
have
like
a
format
that
is
specified
and
is
its
own
NPM
packages.
I
can
be
format,
and
things
like
this
so.
F
F
A
E
B
A
E
There
is
a
very
lengthy
document
into
a
new
documentation
that
describes
all
of
the
moving
pieces
of
that
of
how
to
use
widgets
in
web
context
in
the
documentation
in
ND
ND
viewer
now
and
give
you
around
with
widgets
I.
Don't
know
she
saw
that
really
yeah.
If
you
go
to
the
end,
if
you
were
ready,
give
your
link,
we
get
list.
A
E
E
A
E
A
A
E
See
actually
place
anything
that
has
mine
type,
something
for
us
Jason
we
with
a
script
tag
of,
and
that
chased
me
in
the
Jason.
If
you,
if
the
notebook
Jason,
you
have
a
key
that
he
work
I
mean
in
the
output
area.
You
have
a
key
that
he's
off
a
certain
man,
type
plus
Jason.
It's
going
to
replace
that
with
a
crypt
of
the
correct
mime
type
with
the
same
content.
A
F
The
demeaning
is
Jupiter
lab
/
notebook,
that's
what
Paulson
that
too,
however,
so
that
the
last
thing
is
a
couple
of
us
are
meeting
together
in
the
of
the
schema
and
and
working
together,
so
Paul
tile
in
etc,
and
iron
and
Sylvan
are
talking
about
next
generation
thing.
Actually,
it's
going
on
right
now
we
stepped
out
of
that
meeting
in
order
to
come
here
about
which
of
models
moving
forward,
how
we
might
work
together
with
line.
E
The
new
look
has
sent,
he
was
probably
saved
and
that's
why
it
doesn't
even
work.
I
have
another
name,
we
just
event
works
yeah,
so
attorney
that
the
notebook
was
corrupted.
So
if
you
check
the
second
links,
you
see
the
widgets
working,
so
I
probably
need
to
reset
that
notebook
before
you
support
in
the
end,
you
see
that
the
JavaScript
burnings
work
for
the
spider.
E
F
A
A
Does
the
tooltip
stay
open
as
the
user
continues
to
type
there's
the
tooltip
in
the
classic.
Notebook
has
a
couple
different
modes
and
I.
Don't
it
doesn't
sound
like
we
need
to
implement
all
those
modes?
We
need
to
have
sort
of
one
general
purpose
mode
and
so
at
least
wanted
to
bring
that
up
and
I'll
post
a
link
to
that
and
the
hack
pad.
But
there's
an
issue
open
and
a
lengthy
discussion,
I
think
there's
more
or.
A
C
A
And
then
I
did
see,
there's
a
Matthias
found
a
bug
with
triggering
the
tab,
completion
scrolling,
the
notebook
page
and
the
example
that
he
showed
was
using
a
cell
that
was
larger
than
the
panel,
and
this
tends
to
be
an
issue
that
we
run
into
a
lot
with
classic
notebook
and
we've
had
to
had
to
sort
of
spend
a
lot.
Do
a
lot
of
work
on
that
particular
edge
case
and
focus
related
issues.
And
so,
as
we
test
different
things
in
the
notebook,
it's
something
to
keep
in
mind.