►
From YouTube: JupyterLab Dev Meeting, July 1, 2016
Description
Meeting of the IPython/Jupyter development team.
Meeting Notes: https://jupyter.hackpad.com/JupyterLab-Weekly-Meetings-UUJ3gIQ3iBS
B
Was
still
needed,
never
right,
let's
start
off
with
the
status
reports
and
then
anything
we
need
to
talk
about
so
I've
been
working,
a
lot
on
the
I
pie,
widgets,
Silva
and
I
sat
down
and
reviewed
it,
and
we
are
this
close
to
merging
it
in
and
then
I'll
work
on
getting
the
widgets
working
with
jupiter
lab
so
I.
One
thing
I
think
we
should
talk
about
is
Steve.
B
C
I
have
been
working
all
week
on
refactoring
testing
and
changing
the
behavior
of
text.
Completion
and
the
widget
is
a
lot
more
hardened
now
than
before.
A
lot
of
the
edge
cases
are
now
covered.
C
Almost
all
the
code
is
covered,
but
I
just
did
a
big
refactor
yesterday
that
will
require
rewriting
or
adding
some
new
tests,
not
rewriting
some.
There
was
an
outstanding
issue
open
about
what
the
behavior
of
the
tab
key
should
be,
and
now
we
have
the
first
time
you
tab
all
if
it
happens
to
be
the
case
that
every
item
in
every
choice
of
tab
completion
is
begins
with
a
common
substring.
Then
it
fills
in
your
text
up
to
that
substring.
If
you
should
hit
tab
again.
Whatever
item
you
have
selected
becomes
quarks
and.
C
There
there
are
a
couple
other
behaviors
that
are
going
to
change.
Is
that
the
completion
which
is
going
to
get
an
anchor
element
before
we
use
have
this
reference
semantic
parent
element,
but
that's
not
actually
what
we
care
about.
All
we
care
about
is
the
scrollable
node,
which
we
want
to
anchor
it
to,
and
so
I've
change.
The
name
of
that
I've
changed
the
behavior
that
and
I'm
working
on
tying
the
scroll
event
to
that
anchor
scroll
top
to
make
sure
you
can
scroll
and
keep
the
completion
widget
up.
C
There
were
some
edge
cases
that
are
kind
of
maybe
tedious
to
get
into
now,
but
anyway,
there
were
some
things
that
have
been
resolved.
If
you're
curious,
you
can
look
at
the
PR,
that's
outstanding
one.
One
thing
that
came
up
today,
actually
that
Steven
I
spent
a
while
discussing-
and
this
is
not
for
immediate
implementation,
either
on
the
client
or
on
the
back
end.
C
But
one
thing
that
came
up
is
that
suppose:
I
am
in
a
cell
and
I
say
from
I
pie,
widgets,
important
slider
and
then
in
that
same
cell
I,
go
to
the
next
line
and
say
x,
equals
in
slide.
I
want
to
create
a
new
instance
of
an
in
slider
right
now.
What
the
current
notebook
does
is
in
addition
to
getting
Colonel
API
responses
for
completions.
It
also
manually
token
eyeses
everything
that's
in
the
code
editor,
and
so
it
might
actually
fill
that
for
you,
because
it'll
know
it's
an
in
slider.
C
That's
one
of
the
strings
you
happen
to
have
in
there,
but
the
colonel
doesn't
actually
do
that
and
the
colonel
could
do
that
because
you
send
the
whole
text
of
the
editor
to
the
colonel
for
completion.
So
at
least
there's
nothing
in
theory,
stopping
the
colonel
from
basically
compiling
that
in
telling
you
what
you
can
do
with
it.
The
same
way
my
text
editor
when
I
import
something,
I
don't
really
go
through
a
compile
phase.
It
automatically
lets
me
use
it.
C
So,
as
a
user
of
an
ide,
I
would
have
thought
that
was
going
to
work
and
that's
what
I
thought.
That's
why
I
thought
the
colonel
was
asking
for
multi-line
code
inputs,
but
it
turns
out
it's
not
really.
It's
not
given
that
back.
So
that's
an
enhancement
that
could
go
into
the
kernel
and
it
would
be
totally
API
backward
compatible.
Nobody
would
notice
because
it
would
just
be
slightly
more
options,
but
it's
gonna
require
some
work.
So
that's
that's
one
thing
the
Bryan
you're
on
mute.
If
you
wanna
yeah.
C
C
The
colonel
does
now
that's
something
that
I'm
gonna
post
an
issue
we
can
solve
at
some
point
in
the
future,
or
we
can
decide
that
the
colonel
is
going
to
be
more
supportive
or
whatever
we
can
come
up
with
with
some
strategy.
What
the
current
notebook
does
is
it's
actually
manually
going
through
the
code,
editor
instance
and
using
this
method
called
get
token
at
and
it
does
each
line
multiple.
It
hits
each
line
and
basically
builds
a
list
of
tokens
itself,
and
then
it
merges
that
with
or
removes
anything
in
that
list.
C
That
is
in
the
API
response
as
well
and
then
serves
the
two
separately
with
the
API
responses
on
top,
so
that
you'll
intuitively
know
those
are
from
the
server
and
the
bottom
ones
are
from
the
client.
Now
we
can
do
something
like
that
and
I.
Don't
I,
don't
think
that's
a
bad
idea,
but
I
do
think
that
actually,
if
you
want
to
feel
like
you're
in
an
IDE,
your
kernel
should
be
giving
you
all
the
things
that
are
available
to
you
whenever
they
become
available
to
you.
D
C
Cool,
that's
that's
useful
to
know,
and
now
we
had
some
cake
or
having
both
that's
the
case,
we're
having
both
and
in
hand
and
on
the
carpet
ID
and
also
having
the
tokens
on
a
front
end.
Now
one
thing
that
I
discovered
looking
at
the
code
mirror
code
is
there
is
a
slightly
less
inefficient
way
of
getting
all
the
tokens.
You
can
get
all
the
tokens
for
a
line
instead
of
getting
individual
tokens.
C
The
way
the
notebook
code
currently
does
so
so
there's
room
for
making
that
slightly
more
efficient,
but
we're
talking
about
an
inherently
inefficient
thing,
and
that
may
be
okay.
That
might
be
all
right
with
us,
or
it
might
turn
out
to
be
the
case
that
once
we
have
a
kernel,
that's
more
productive
in
what
it
gives
back
for
completion.
C
C
D
May
be
right
that
we
may
I'm
just
trying
to
think
exactly
the
the
completion
from
other
cells
may
be
coming
through.
The
colonel
actually
I'd
have
to
look
at
that,
and
so
may
it
may
be
that
what
you're
saying
that
we
can
implement
the
colonel
token
recognition
in
a
given
cell
and
we'd
be
covered
right.
C
Yeah,
actually
it's
it's
more
than
just
hope
token.
Recognition,
because,
if
I
create
a
new
in
slider,
I
really
want
to
know
the
methods
of
that.
So
it's
not
just
enough
to
know
that
there's
a
token
called
in
slider.
The
colonel
actually
has
to
have
compiled
it
to
be
able
to
give
me
more
than
just
the
tokens
available.
It
has
to
import
in
slider
yeah.
D
And
just
it
also
a
bit
of
background,
there
has
been
pretty
strong
interest
in
improving
the
a
two
things:
one,
the
message:
spec
for
tab
completions
to
allow
kernels
to
provide
more
additional
information
such
as
types,
so
there's
interested
interest
in
improving
the
message
back
part
of
it
and
then
obviously
improving
the
default
python
kernels
implementation
of
that
to
provide
that
additional
information
will
Whitney
last
summer
started
to
work
on
that
I'm.
Guessing
we
I
think
we
have
an
enhancement
proposal.
You
need
somewhere
that
we
could
could
bring
up
to
see,
see
that
stuff.
C
Yeah,
that's
cool.
That
would
be
great
actually
to
have
a
completion
widget,
that's
more
more,
like
here's,
a
method
signature
in
line,
instead
of
just
getting
that
in
tooltips
or
whatever
and
the
other.
The
other
thing
is
that
these
can
be
done
in
multiple
stages,
and
some
of
them
will
be
totally
backward
compatible.
So
just
getting
the
extra
tokens
or
getting
the
extra
tokens
and
their
sub
methods
or
whatever
that
can
be
done
in
the
backward
compatible
way.
C
It'll
just
be
a
longer
list
of
completions
and
the
metadata
that
you're
talking
about
obviously
would
require
some
sort
of
release
schedule
to
get.
You
know
it
would
it
would
deprecated
some
API,
so
that
would
be
different,
but
there
are.
There
are
multiple
tiers
that
we
can
hit
this
app
without
causing
disruption.
Okay,.
C
So,
but
for
immediate
release,
I'm
going
to
finish
up
this
PR
without
any
sort
of
clone
inside
tokenizing
and
then
create
a
separate
issue
for
that,
because
the
completion
widget
has
changed
enough.
That
I
want
people
to
start
using
it.
I
want
to
get
it
out
of
the
PR,
and
this
is
a
story.
That's
containable
as
a
totally
separate
issue.
So
that's
that's.
How
I'm
tackling
it?
Ok
and
a.
C
C
F
Yeah,
so
my
work
update,
isn't
much
different
than
backe,
still
taking
a
few
days
off
and
working
on
making
the
finishing
up
the
final
bit
of
bring
bring
the
last
widget,
which
is
Dom
panel
into
the
mono
phosphor
mono
rego,
so
that
works
still
ongoing,
hopefully
have
that
done
today.
My
ultimate
goal,
though,
is
to
have
that
mono,
repo,
ready
and
Jupiter
lab
running
on
the
new
pontiff
repo
by
the
time
Syfy
hits
so
I'll
work
with
Steve
on
that
next
week.
B
F
Going
to
put
that
in
the
Montori
platoon
that'll
be
part
of
that,
so
there
will
be
discussions
at
some
discussions
about
that.
That
will
happen
next
week
that
I'll
just
probe,
you
guys,
are
not
as
we
need
eye
on
them
about
unifying
services
and
extensions
and
what
we
want
to
do
about
the
type
safety
of
that
speed
brought
in
Steve
brought
up
a
point
about
yesterday
about
wanting
to
use
dependency,
injection,
essentially
extension
mechanism
and
it's
a
dependency
system
for
reified
instances
of
generic
classes.
F
So
let's
say
you
have
some
class,
that's
generic
over
some
type
and
you
want
something
that
provides
a
concrete
version
of
that
so
fully
ratified,
like
a
vector
of
its,
for
example,
just
an
example
of
a
rate
buy
generic,
and
so
how
do
we
spell
that
out
until
then?
Something
that
will
address
and
put
full
brain
cycles
into
that?
Yet
so
the
solution
may
be
trivial,
it
may
be
not
reveal,
will
flush
that
out
next
week,
cool.
E
Any
given
time
also
look
to
add
so.
D
E
E
D
E
And
the
other
one
was
just
a
changes
to
phosphide
to
just
temporarily
allow
us
to
put
content
in
the
top
area,
or
that
was
sort
of
stubbed
out
before
it
wasn't
implemented,
and
then
also
PR
against
super
lab.
To
allow
us
to
add
things.
Similarly,
to
with
the
same
API
is
adding
things
the
left
panel,
where
you
can
provide
a
rank
to
add
things
to
the
top
level
menu.
D
And
then
Chris,
so
right
now
that
API
you
actually
pass
menu
items
in.
Is
that
correct
and
those
in
those
menu
items
which
are
we
talking
about
amount
of
phosphor
menus,
no
Chris
sold
on
three
E
and
I?
Guess
I
just
want
to
understand
sort
of
how
that
API
will
change
when
we
could
go
to
the
mono
repo,
okay.
F
We
take
the
opportunity
to
unify
menu
items,
the
menus
with
commands,
and
so
the
deland
menu
items
to
a
menu
button
menu
item
is
basically
just
a
closure
over
a
command
in
the
arguments
for
a
command
and
then
the
menu
looks
up
all
that
data
for
drug
to
render
the
render
the
video
item
from
the
from
the
command
registry
and
stuff
that
goes
with
the
commands,
and
so
in
the
simplest
case.
The
menu
item
is
nothing
but
a
command
ID
and
all
the
command
takes
no
are
so.
Okay
and.
D
F
One
thing
I
did
come
at
Bryan
you're
on
your
post
about
the
the
Jupiter
icon
going
in
there.
So
the
menu
item
sells
the
thing
that's
rendered
in
a
menu
bar
like
the
Texas
Rangers
in
the
minibar
there's,
actually
an
icon
node
there.
So
we
can
set
an
icon
like
let's
say
the
file
menu
is
the
leftmost
one.
We
can
set
an
icon
on
that.
F
F
F
D
E
E
Lot,
okay
and
then
the
last
part
was.
I
was
looking
at
ways
to
basically
circumvent
mpm
and
not
require
and
users
to
have
em
p.m.
if
possible,
when
adding
new
third-party
extensions,
for
example,
BQ
plot,
and
to
see
if
that
was
even
possible.
The
first
thing,
I
tried
was
there's
a
dll
plugin
for
web
pack,
where
you
create
the
dll
on
one
end
and
then
there's
a
deal,
a
reference
on
the
other.
But
what
I
found
out
is
that
that
is
brittle
to
the
point
of
even
in
a
patch
release.
E
You
could
break
that
API,
so
you
did
you'd
have
extensions
that
would
we're
tied
to
an
exact
patch
release
of
Jupiter
lab.
So
that's
a
note
non-starter,
oh,
but
what
we,
what
we
might
be
able
to
do
is
to
provide
UMD
bundles
for
the
things
that
we
want
to
share
and
then
at
least
be
able
to
within
a
patch
release
remain
in
lockstep.
So
we
can
use
some
or
at
least
to
that
level
between
the
Jupiter
labcorp
and
the
extension.
So
that's
what
I'll
be
exploring
some
next
week.
A
B
Jason,
did
you
already
give
an
update?
I
did
we're
working
on
widgets
disclose
to
merging
so
when
I
sat
down
and
I
get
together
turns
out
the
code
with
pheromone
I
forward
funny
better
now
and
there's
several
iterations
that
we
didn't
want
to
put
in
this
sort
of
big
PR
a
living
effects,
script
that
I'll
go
to
follow
pretty
quickly
after
you
could
be
sex
with
the
RN
C.
B
So
can
you
just
walk
us
through
this
real
quick,
a
little
bit
more
carefully,
so
I'll
constructed
the
bundle
and
distribute
that
with
my
NPM
package
and
then
somehow
Seljuk
their
lab
load?
This
requires
a
UMD
bundle,
so.
E
Fleshed
out,
but
yes,
and
you
right
now,
Jupiter
jeaious,
widgets
and
do
purchase
services
are
both
shipping
bundles
with
them
on
their
AMD.
But
it
doesn't
matter
Andy
for
what
come
out
of
this
web
packs
perfected,
but
as
long
as
it's
in
a
format
that
web
pack
can
consume
as
an
external
arm
and
that
the
show
that
the
actual
clothes
in
the
loop
I
haven't
done
yet,
but
I
believe
it's
possible
how
painful
that
is
remains
to
be
seen,
live
hello,
everybody
time
from.
H
H
B
Cool
okay,
I
think
we're
the
Cal
Poly
Carol
Brian.
D
Come
yeah
I
don't
know
just
did
the
Cal
Poly
teams
want
to
give
a
quick
update
on
what
your
you've
been
working
on.
You'll
definitely
have
to
get
close
to
the
microphone.
Maybe
just
go
up
and
grab
it
and
and
chat
just
yeah.
You
can
just
then,
through
the
like
a
quick
outline
of
the
things
you're
working
on
right
now.
D
I
Yeah
right
now
we're
just
kind
of
doing
a
couple:
little
design,
ketchups
onda,
some
of
the
older
widgets
like
well.
We
were
going
to
put
a
menu
bar
into
that
editor
widget,
but
now
there's
a
better
option.
You
know
where
we
can
have
an
into
global
menu
bar,
so
I
think
today,
I'm
gonna
go
about
adding
a
few
many
bars
or
many
of
our
items
that
we
can
put
up
there
in
that
global
menu
bars.
I
It's
kind
of
like
a
walkthrough
just
it
has
most
of
the
same
information
on
the
current
dog
page.
However,
it's
a
little
bit
more
like
visual
and
it'll
actually
show
users.
Oh,
this
is
how
you
actually
drag.
You
know
a
tab
and
place
it
another
like
framing
frightened.
Oh
there's,
more
visual
stuff
like
that
and.
D
Then
did
you
I,
don't
know
students
who
had
dropped
off
for
a
second
there,
but
we
could
win
headed
merged,
Stevens,
widget,
tracking
PR
and
then
we're
going
to
merge
the
menu
PR
right
after
this
there's
a
few
merge
conflicts
that
steven
has
to
clean
up
first.
So,
though,
both
of
those
will
be
a
master
and
go
you'll
be
ready
to
go
on
the
menu
than
widget
tracking
cool.
I
A
I
H
I
G
E
What
do
you
mean?
We
don't
actually
have
jquery
on
the
page,
yet
Jason
good
might
I
had
given
Elliot
some
code
to
add
it,
but
it's
not
actually
on
master.
Yet,
ok.
G
D
But
did
you
did
I
see
that
there
is
a
bug
in
the
terminal
stuff
right
now,
but
did
you
work
out
yet
just
trying
302.
E
The
the
terminal
was
the
it's.
The
terminal
underlying
xterm
widget
was
not
widget,
but
I
object
was
being
created
after
the
WebSocket
was
open,
and
so
it
was
know
when
they
were
trying
to
access
it.
So
there's
a
PR
open
right
now
that
cleans
that
up,
so
we
create
the
widget,
but
don't
don't
populate
it
or
interact
with
the
dimensions
of
it
until
such
a
time
as
actually
the
web
sockets
open.
E
D
So,
like
you
can
set
font
size
beforehand,
but
it's
not
gonna
take
take
effect
until
it
all
triggers
right.
Ok,
yeah,
I
was,
I
was
very
confused.
I
look
at
looked
at
the
code
with
turn
treat
and
was
like
I
think
you're
setting
it
correctly.
It
looks
like
that's
a
little
odd
great
thanks.
So
much
anything
else.
The
students
want
to
mention.
That's.
A
A
D
H
E
Yeah,
if
you
could
open
an
issue
on
Jupiter
lab
I,
think
that's
a
good
place
for
them.
Yeah.
B
At
least
yep,
yes,
Carol,
you
want
to.
J
Give
an
update,
yeah
thanks
again
for
all
the
help
that
the
team
is
giving
to
the
Cal
Poly
interns.
It's
much
appreciated.
I
am
gonna,
probably
towards
the
end
of
next
week.
Work
on
still
fleshing
out
that
tutorial.