►
From YouTube: JupyterLab Weekly Dev Meeting, February 3, 2017
Description
Meeting of the JupyterLab development team, February 3, 2017
Meeting Notes: https://jupyter.hackpad.com/JupyterLabNotebook-Weekly-Meetings-UUJ3gIQ3iBS
A
Hello
Internet:
this
is
the
februari
third
Jupiter
labs
and
deeper
notebook
that
meeting,
and
today
we
have
a
smaller
kuya
normal
and
you
don't
yet
have
our
hack,
that's
I'm,
not
sure
who
should
go
first.
So
I
will
nominate
steve
and
I
actually
know
what
we
normally
do.
P.M.
/
sewing
so
Jamie
of
one.
You
guys
have
any
administrative
things
to
talk
about
I.
C
I've
been
in
serious
teaching
and
administrative
bureaucracy
type
stuff
this
week,
so
it's
very
I've
gotten
a
lot
done,
but
nothing
that
feels
productive.
Let's
put
it
that
way.
A
I
think
I'll
go
I've
mostly
been
thinking
about
completer
this
week.
Both
there
were
some
bugs
I,
think,
there's
still
one
outstanding
bug
and
also
just
the
notion
of
how
do
we
add
functionality
to
Jupiter
lab.
If
we're
thinking
as
though
we
are
3rd
party
plugins
others,
instead
of
thinking
as
those
people
as
creators,
so
one
of
the
things
that
came
out
of
building
the
toolkit
food
that
was
having
it
be
completely
isolated
to
a
plug-in
and
so
I'm
still
I
was
actually
going
to
try
to
do
the
same
thing
with
completer.
A
But
I
saw
like
lower
hanging.
Fruit
was
the
inspection
we'll
just
so
right
now,
both
the
milk
buggin
and
the
console
plugin
listeners
on
their
own
individual
instance
trackers,
and
they
were
basically
manually
swapping
out
the
source
of
the
infection
handler.
But
really,
if
you
think
about
who
might
have
created
the
inspection
handler
instead
of
us,
it
might
have
been
just
some
third-party
plug
in
a
1.
Go
modifying
notebook
and
counsel.
So
I
moved
that
manual
swapping
in
inspection.
E
There
are
there's
a
very
simple
way
to
instantiate
a
selector
and
based
on
what
was
in
the
previous
notebooks
and
that
you
give
it
the
key
the
metadata
key
you're
looking
for
and
a
set
of
auction
name
value
pairs,
and
so
it
could
create
a
selector
widget
for
you
and,
as
you
change
the
drop
down,
it
keeps
the
metadata
and
vice
versa.
So
we've
got
that
done
done
for
slideshow
and
any
convert
pulse.
E
F
I'm
really
quick,
I've
been
working
on
widgets.
We
put
our
new
widgets
data
in
front
of
some
users
and
immediately
got
some
feedback
like
I
tried
to
do
this
and
it
didn't
work,
ok
back
to
the
drawing
board,
so
then
plugging
ahead,
so
impulsive
and
plugging
head
on
with
it
fixing
some
issues
kind
of
really
try
to
push
ahead
on
issues
for
six
point
o
other
one
big
thing
for
widgets
was
I,
still
have
a
question
of
notebook
4.4
released
as
a
back
port
for
the
output
widget.
F
That
makes
widgets
and
interact
and
cetera
work
in
notebook
for
X,
and
we
could
release
a
4.4
that
just
has
that
one
back
for
it.
It's
not
an
issue
right
now,
because
six
point
0
is
not
released,
but
once
which
is
six
point,
0
gets
really
either
addict.
We
should
release
a
notebook
4.4
or,
if
I
point
O
has
been
out
for
a
long
time.
C
F
Been
working
with
him
on
that
I
didn't
know
if
he's
kept
doing
it
I
just
changed
a
bunch
of
stuff
this
morning,
because
I
realized
that
setting
the
width
on
a
widget
didn't
work,
and
so
that
might
have
affected
a
bunch
of
the
wits
and
stuff
like
that.
The
way
I
three
architects
a'junk,
so
I'll
get
August
together
with
him
and
and
double
check.
What
he's
doing
and
yeah
tell
them
to
update
to
the
newest
version
which
which
might
change
from
the
cff
thirties
day.
Okay,.
C
F
A
G
A
A
Sophistication
we
need
to
offer
because
editing,
arbitrary
data
structures.
You
know
you
could
have
things
that
generate
forms
or
you
know,
there's
a
lot
of
ways
of
doing
it.
Basically
or
you
could
just
give
them
a
code
editor
and
you
know
pretty
print
it
and
say:
go
to
town,
so
I
I
don't
know
but
see
if
you
probably
have
more
coherent
than
this
one.
What
are
you
thinking
was
that
one
else
thank
you.
E
I'll
just
lay
out
what
Darren
and
I
had
proposed
the
other
day
and
that
we
stick
with
a
raw
text
area
but
make
it
clearer
in
terms
of
what
state
were
in
so
that,
if
we're
focused
in
a
text
area
that
is
considered
live
mode
in
it,
so
it
doesn't
change
the
data
structure
under
us.
Even
if
the
actual
underlying
structure
changes
and
at
the
point
you
make
a
change
is
when
it's
considered
40.
We
have
some
kind
of
indication
of
that.
Some
buttons
come
up
that
either
you
can
accept
or
revert
at
that
point.
E
If
you,
if
you
haven't,
made
any
changes,
so
you
just
focus
in
the
cell
and
then
you
blur
the
cell
and
leave.
Then
it
should
update
tool
to
the
underlying
state.
If
you
leave
in
the
midst
of
editing,
it
should
stay
in
that
dirty
state.
If
you
cancel
editing,
it
should
be
as
if
it
were
just
revert
to
the
underlying
data
structure
and
then
the
point
where
it
gets
messy
is,
if
you
commit,
then
you
have
to
do
is
merge
of
so
whatever.
Presumably
whatever
the
user
has
entered,
would
kick
priority.
E
At
that
point
she
emerged
the
keys,
but
allowing
what
the
users
entered
to
check
priority.
That
assumes
that
the
users
actually
entered
valid
JSON.
So
we
have
the
first
half
developing.
The
JSON
shows
some
kind
of
error
indication
if
it's
not
valid
and
then
once
they
finally
have
committed
valid
JSON,
then
we
do
the
deeper
and
everything's
back
to
this
snapshot
state.
So
that
was
that
was
our
thoughts
of
sort
of
a
cheap
wave
of
we
miss
editing
of
data
that
can
change
from
underneath
you
that's
arbitrary.
E
C
Gone
wrong:
yeah.
D
F
C
11,
actually,
one
thing
that
might
help
in
this
is
it
wouldn't
be
a
modal
dialogue,
but
it
would
be
one
make
it
slightly
more
modal
and
it
might
help
would
be
to
offer
to
allow
cell
tool
plugins
to
be
collapsed,
so
they're
not
seen,
and
for
cell
toolbar
or
cell
tool
plugins
to
specify
whether
they
should
be
open.
You
know
open
or
or
collapse
by
default,
and
it's
how
you
can
just
share
my
screen.
Real,
quick
I'll
show
you
what
this
ends
up.
Looking
like.
D
C
C
Now,
in
this
case,
when
you
click
on
them,
they
open,
although
you
can
see
still
the
shadows,
inner
shadows
and
Gaussian
blur
sections
are
still
closed
and
what
I'm
wondering
is,
if
that
type
of
sort
of
collapse
and
expand,
interaction
could
be
used
to
make
the
cell
metadata
the
raw
sell
editing
of
that
metadata.
Morimoto,
basically
going
this.
A
So
so
the
life
cycle
that
these
describing
is,
what
do
you
do
then?
Do
you
overwrite
the
192
to
the
new
value,
or
do
you
leave
it
at
192,
because
you,
the
user,
have
decided
to
focus
on
that,
meaning
you
might
have
a
value
you
want
to
put
in
a
superseded
anything
else.
So
modal
or
not
that's,
actually
a
possibility,
even
in
the
current
notebook,
it
would
be
a
possibility.
It's
just
a
it
comes
more
to
the
fore
if
both
are
actually
even
still
visible
and
and
that.
C
And
so
it's
some
senses.
It
boils
down
to
the
fact
that
the
metadata
is
not
a
fully
live
objects
that
we
can
real
time
sink
I
mean
it
for
objects
where
we
have
actual
full-blown
models.
It's
not
a
problem
to
have
multiple
views
connected
to
that
model,
and
so
it
seems
like
that
right
I
mean
it's.
Every
single
metadata
field,
all
the
way
down
its
hierarchy,
was
a
real
model.
Every
view
would
be
often
synced
to
the
view,
and
so
it's
the
fact
that
parts
of
the
metadata
are
opaque.
Blobs
that
may
change
well,.
E
C
C
You
know
proper
full-blown,
codemirror
support
for
multiple
cursors
and
real-time
models,
and
all
that
like
that
we,
but
the
challenge,
though,
is
if
we
do
that
one
that
makes
in
job
a
lot
more
difficult
because
he's
going
to
have
to
real
time
to
sync
recursively,
all
the
way
down
that
metadata
structure
and
we'll
have
to
do
that
everywhere.
In
Jupiter
lab
it's
I,
I'm
not
I'm
just
trying
to
distill
what
the
source
of
the
problem
is,
I'm
not
proposing
that
we
actually
like
make
it
a
full-blown
model,
but
it
seems
like
that.
A
Yeah,
it
is,
but
I
think,
there's
a
simpler
way
of
thinking
about
it.
So
when
I
opened
my
text
editor,
it
treats
my
file
like
a
just
blob
of
just
a
massive
stream
right
and
then,
if
I
do
like
a
get
revert
or
something
and
I
come
back
and
I
have
unchanged
things
in
my
buffer.
My
text
editor
basically
gives
me
a
choice.
A
You
want
me
to
give
you
what
the
file
actually
is
or
do
you
want
to
just
continue,
and
we
can
do
that,
basically
just
ignoring
the
complexity
of
actual
real
time
collaboration,
because
this
is
a
black
box.
That's
change,
it's
not
a
real
model,
so
I
think
there's
like
a
middle
ground,
where
we
can
basically
acknowledge
that
this
thing
has
been
pulled
out
from
under
you
and
let
you
make
a
decision
yourself.
A
A
C
C
F
F
E
F
F
A
If
keys,
get
deleted
and
you're
not
modifying
them,
they
automatically
delete
in
the
single
weird
is
of
you,
editing
the
one
that
happens
to
have
been
deleted
or
modified.
Then
it
gives
you
an
indication,
but
otherwise
it's
just
a
regular
text
editor
with
a
regular
toolbar
and
one
of
the
things
in
the
toolbars
to
drop
down
for
all
the
top
level.
Things
yeah.
F
I
was
kind
of
suggest
looking
at
what
Visual
Studio
code
is
doing
with
their
settings.
They've
got
a
really
powerful
editor
for
the
JSON
settings.
That
essentially,
is
just
JSON,
but
there's
tons
of
tools
in
there
to
make
it
easy
to
navigate
an
easy
to
collapseexpand
sections
easy
to
they've
got
choices.
So
if
you
completion
and
things
like
that
with
the
JSON-
that's
smart.
So
so
it
knows
the
values
that
need
to
be
in
a
particular
setting
and
it'll
give
you
a
completion
or
a
drop
down.
A
Is
the
one
the
one
thing
that
we
discussed
a
little
bit
before
the
meeting
was
just
a
complexity
of
writing
your
own
gift
on
editor
or
there's
a
million
libraries
that
do
it,
but
they
all
come
a
trade-off.
You
know
we
would
have
to
basically
figure
out
all
right.
What's
the
interface
that's
ideal
and
then
figure
out
what
level
of
commitment
and
work
it
takes
to
deliver
that,
like
an
arbitrary
data
structure,
editor,
that's
user
friendly,
is
itself
kind
of
a
big
problem,
so
good
yeah.
D
F
Here's
Visual
Studio
code,
I,
think
I've
Jupiter
lab
open
here,
but
I
go
to
a
setting
and
here's
the
settings
interface.
On
the
left
hand,
side
is
all
the
default
settings,
so
this
is
sort
of
wish
system-wide
configuration
on
the
right
hand.
Side
is
my
user
settings
and
they've
also
got
workspace
and
language
now
settings,
but
you
can
see,
for
example,
there's
sections
collapsible
sections.
F
F
F
I,
don't
know
if
it
was
pulling
it
out
of
the
comment
or
if
it
was
keema
I
wouldn't
be
surprised
if
it
was
a
schema
that
they
were
trying
to
visit
that
they
were
satisfying
here.
But
the
point
is
you
can
make
a
JSON
editor
I've
been
really
impressed
with
how
friendly
they've
made
this
JSON
editor
to
edit.
Just
what
normally
is
just
a
huge
text,
file,
settings
and
I
guess
up
here?
Workspace
I
can
do
workspace,
setting
versus
user
settings
yeah
anyway.
A
F
A
A
C
C
E
C
F
Yes,
yes,
you
might
disagree
with
that.
Well,
the
point
is
that
we
might
have
to
rethink
a
lot
of
this
once
in
stuff,
as
easing
in
suffers
iterates
and
evolves.
So
so
do
we
punt
now
or
do
we
do
some
level
of
thinking
that
we
may
have
to
redo
later
and
we
have
it?
How
much
do
you
have
for
a
sidebar?
It
sounds
being
able
to
select
multiple
cells
and
change.
The
metadata
on
multiple
thousands
so
sounds
really
powerful,
but
there's
a
lot
of
corner
cases
in
there
as
well.
Yeah.
A
Think
this
is
really
thing
where
no
matter
what
quantity
of
things
you
select
when
you
say
get
info,
it
shows
you
the
subset
they
have
in
common
that
you
could
modify.
So
if
I
saw
it
like
all
of
the
songs
by
an
artist,
I
could
set
them
all
to
be
the
same
thing.
But
it's
not
going
to
fill
out
the
values
for
the
things
that
are
heterogeneous.
It'll
only
fill
out
the
values
for
things
that
they
have
in
common,
but
apples
on
that
for
20
years.
So.
F
F
C
Only
I
think
I
would
probably
prefer
that
basically
what
you're,
proposing
Steve,
but
just
not
modal
and
reason
I,
would
say
that
that
I
want
to
reserve
that
modal
interaction
or
when
the
data
changes
from
underneath
the
user
and
asking
them
hey.
Do
you
want
to
you
want
to
force
your
changes
or
accept
those
other
changes?
The.
F
Reason
why
I
don't
like
modal
is
because
we're,
unlike
the
notebook
where,
essentially
it
is
a
modal
interface,
one
notebook
at
a
time
here
we're
dealing
with
multiple
notebooks
and,
and
we
really
want
to
support
being
it
be
able
to
switch
between
multiple
notebooks.
So
what
happens
in
your
sidebar
editor
if
I
switch
notebook
to
the
update?
Yes,
yeah?
Okay,
so
that's
really
powerful
to
be
able
to
copy
settings
for
muscle
in
another
notebook
to
the
current
notebook
to
be
able
to
switch.
F
F
A
It
sounds
like
the
original
sort
of
life
cycle
that
that
deep
described
is
covered,
our
basic
use
case
and
and
when
we
have
some
like
real
real
time
stuff
going
on,
might
consider
all
right.
How
do
we
make
this
more
nuanced,
but
at
least
was
Lucy's
discuss
is
not
just
feature
parity.
It's
actually
slightly
nicer
actually
sufficiently
nice.
When
we
talk
about
switching
between
multiple
notebooks,
but
it's
a
yeah
I.
Imagine
it's
a
pretty
good
place
to
introduce
the
feature,
and
then
we
can,
you
know,
modify
it
as
we
go
forward.
Yeah.
F
E
Yeah
the
default
plug-in
that
is
provided
by
the
notebook
bullion
final
is
just
assembling
these
things
together,
but
you,
a
third
party,
can
come
in
and
create
a
new
soul
tool
that
is
even
the
baseball
to
lure
from
this
selector
subclass.
That
makes
that
part
in
here
and
then
just
add
it
with
a
rank
and
will
show
up
in
the
appropriate
location
in
the
soul
tools.
Okay,
can
those
couple.
F
E
A
F
Have
a
different
plugins,
my
own
different
keys
in
the
settings
dictionary
because
I
each
plugin
has
its
own
collection
of
settings,
so
it
might
provide
its
own
UI
for
editing
that
and
but
you
still
want
to
edit
it
at
a
JSON
level.
So
it's
easy
to
copy
and
paste
and
just
deal
with
things
in
a
text.
Format,
I.
E
C
And
I
think
one
I
think
the
type
of
thing
we
could
do
if
we
want
to
embrace
the
file
approach
on
a
per
plugin
basis
is
have
Jason
schema
for
these
config
things
that
allows
us
to
validate
either
as
the
user
types
or
when
they
favor,
but
I
think
that
would
help.
We
could
also
offer
in
principle
completions
based
on
that
schema,
so
I
think,
because
the
having
jisang
config
schema
is
going
to
go
long
ways
towards
enabling
us
to
offer
short
of
the
vs
code
style.
E
Know
that
is,
you
know,
there's
an
open
issue
for
that.
It's
not
marked
data
at
one
point,
oh
right
now
that
that
one
could
either
be
another
side
bar
that
shows
up
temporarily
and
then
has
an
X
on
it
to
hold
it.
The
presents,
well,
I,
don't
know,
could
do
we
want
to
keep
populating
these
cyborgs
like
models
are
bad,
but
how
do
we
have
this
concept
of
a
like?
A
transient,
sidebar
yeah.
C
A
Just
had
to
be
ability
to
help
I
mean
we
could
it's
trivial
that
commands
the
clothes,
even
if
we
don't
have
a
nice
to
put
an
X
or
we
could
just
come
instantly
start
putting
toolbars
on
top
of
sidebar
that
holds
or
whatever,
but
in
terms
of
closing
them
like
it's
fairly
trivial.
To
do,
and
at
least
putting
it
into
menus
is
pretty
easy
to
do.
C
A
A
C
So
we've
been
talking
a
min
and
I
Jeff
and
Carol
been
talking
some
about
hub
share,
so
I'm
sure
it's
going
to
be
the
Jupiter
hub
service
that
provides
sharing
the
sharing
of
directories
basically
and
the
the
basic
way
to
think
about
it
is
that
you
can
take
a
directory
in
your
stupid
or
notebook
server
and
say:
hey
I
want
to
hear
this
with
people.
It
will
get
zipped
up
and
copied
to
the
hub,
share
service
and
then
made
available
to
those
individuals,
and
they
can
then
say.
F
C
Wow,
you
know
Steve's
share
this
bundle
with
me.
This
directory
I
want
to
just
copy
it
into
this
other
directory
in
my
Jupiter
notebook
server,
and
so
it's
all
copy
based
and
you
put
it
like
a
conceptual
perspective,
it's
equivalent
to
something
like
get
where
you
only
track
a
single
parent.
So
you
directories
know
what
parent
they
came
from,
but
there's
no
merging
big
emergency
eyes
down
to
keep
ours
and
keep
theirs,
and
this
is
this
is
basically
needed
to
match
the
semantics
that
we
need
for
and
be
greater.
C
So
but
from
the
UI
perspective,
the
hub,
like
working
with
hub,
share
in
in
lab
you're,
going
to
need
to
look
at
the
file
browser
and
the
hub
share
panel.
That
shows
what
things
have
been
shared
with
you
or
what
things
you
have
shared
with
other
people,
and
so
you
know
I
think
an
optimal
way
of
handling.
C
It
would
just
be
to
drag
things
back
and
forth
between
those
two
panels,
and-
and
so
this
is
an
example
of
another
plug-in
that
sort
of
works
in
conjunction
with
the
file
browser
and
I
can
think
of
two
other
examples.
One
is
the
launcher
right,
you
sort
of
need
to
look
at
the
launcher.
At
the
same
time,
you
look
at
the
file
browser
and
I
that.
C
Well,
then,
actually
there's
one
other
example
of
this
type
of
panel,
and
that
would
be
like
a
version
control
panel
right,
presumably
I
mean
in
in
Visual
Studio
code,
their
file
browser
and
their
birth
to
get
plug-in
or
both
in
the
left
panel.
But
if
you
can
show
both
of
those
at
the
same
pond,
there's
a
lot
more
that
you
can
do
and
so
I
just
think.
It's
a
general
pattern
that
we're
going
to
run
into,
and
it
would
be
great
to
have
sort
of
a
human
form
way
of
handling
these
things.
C
E
C
A
C
C
Ok,
so,
but
if
if
the
user
can
freely
move
things
back
and
forth
between
the
left
and
right
side
panel
and
actually
X
something
out
that
they're
not
using
I,
think
if
we
did
that,
for
example,
I
I
might
advocate,
I
mean
we'd
have
to
play
with
it,
putting
called
washer
on
the
right
side
panel
so
that
you
could
always
see
it.
At
the
same
time,
you
see
the.
E
C
Mean
that
that
was
it's
on
a
typical
laptop
in
business
continue.
This
would
continue
to
be
an
issue
if
you
have
the
left
and
the
right
side
panels
open
the
the
doc
panel
is
quite
compact
and
you're
sort
of
limited
to
one
wat
one
panel
wide
so
now
on
a
big
screen
having
the
left
of
the
right
is
spectacular.
It's
very
nice.
A
I
don't
know
if
we
can
do
this
right
now,
but
I
don't
see
a
theoretical
ism
as
we
couldn't
have
you
open
the
file
browser,
you
start
dragging
a
file
and
you
hover
over
the
launcher,
tab
and
the
launcher
open,
and
then
you
drop
it
onto
the
big
launcher
that
you
want
so
that
they're
in
the
same
bar,
but
you
can
just
by
dragging
over
something
you
can
get
it
to
activate
like
on
a
Mac.
A
A
Well,
what
if
I
want
to
drag
I
mean
this
is
a
bad
example,
because
we
have
context
menus
but
suppose
I
have
a
file
and
I
want
to
open
in
the
editor
and
I.
Don't
know
that
I
can
right-click.
Can
I
maybe
drag
the
file
over
to
the
launcher
on
the
sidebar
and
then
on
top
of
the
file,
editor
icon
or
something.
A
I
mean
I,
you
know
I'm
not
actually
convinced,
because
people
really
love
to
demo
drag
and
drop,
but
I
don't
see
people
really
using
it
much
in
sideways,
like
people
often
just
refuse
to
believe
that
it's
going
to
work
inside
of
web
fish,
though
I
thought
the
dragon
droppable
features
that
people
just
don't
use
and
I,
wouldn't
be
surprised
if
we
find
if
we
found
in
testing
that
people
just
aren't
doing
it.
Yeah.
C
C
A
I
know
that's
for
the
current
launcher,
though
I
don't
know
if
you've
noticed
it
will
only
ever
show
if
there's
nothing
of
your
doc
panel.
So
if
you
have
a
bunch
of
notebook
soap-
and
it
won't
show
because
already
know
what
you're
doing,
but
if
you
close
everything
when
you
refresh
that
page
is
going
to
show
because
it's
assuming
I
something
should
be
in
the
doc
cam
like.
F
F
E
That
other
yes,
just
some
food
for
thought,
I
looked
at
what
cloud
nine
is
doing
on
this
and
and
when
you
right
click
on
either
side.
You
get
an
option
of
all
available
side
panel
items
to
show
on
a
given
side
and
you
can
toggle
them
there
and
what
we
could
do
is
allow
the
user
to
show
and
hide
on
either
side
from
that
list
and
then,
if
there's
nothing
there
just
have
enough
so
that
you
can
still
right
click
there,
but
minimize
the
space
of
that
side
panel.
So
it's
still
visible
one.
F
C
E
C
D
C
A
F
B
A
We
yeah,
but
I'm
not
usually,
but
that's
bad,
because
you
know
this
is
just
where
it
goes
and
you're
in
the
sidebars.
If
there
were
some
visual
indicator
inside
that
right
click
context
menu
that
shows
what
side
was
on.
I
think
that
would
have
been
sufficient.
Can
you
but.
A
F
C
Okay,
but
I
mean
we
could
have
something
like
a
little
icon
below
all
the
open
tabs
that
pulls
up
at
some
sort
of
men.
You
like
that
I
think
having
some
visual
indicator
is
nice,
it's
more
discoverable,
but
also
we
could
show
that
indicator
when
the
top.
The
sidebar
is
completely
gone.
I
catch
some
little
+
tab
that
sticks
out
or
we.