►
From YouTube: Jupyter/IPython Dev Meeting, July 19, 2016
Description
Meeting of the IPython/Jupyter development team.
Meeting notes: https://jupyter.hackpad.com/July-2016-NIpXzXG3Cmt
A
The
project
was
very
well
represented.
I
want
to
thank
everyone
on
the
team
who
participated
and
and
sort
of
made
up
made
us
proud
and
hopefully
I'll
make
it
all
making
thanks
you.
This
was
this
time
when
I
just
had
a
conflict,
but
anyway
Carol
is
the
first
on
the
agenda,
but
she
can't
join
us
so
we'll
leave
her
notes
will
leave
her
notes.
There,
Oh
Pete,
I,
think
you're
next
up
I
know
if
you
want
to
go
ahead.
B
B
B
D
You
know
yeah,
so
I
think
the
main
item
worth
mentioning
have
I
been
out
on
vacation
for
the
past
week,
so
I'm
getting
back
into
the
speed
of
things,
but
we
started
some
work
a
few
weeks
back
and
we
just
recently
moved
it
on.
There
are,
and
we
point
declarative,
widgets
a
stop
repo
and
in
the
incubator
section
it's
about
it's
a
widget,
it's
a
declarative
widget
that
allows
you
to
tie
a
2-2,
a
data
frame,
a
panda
spark
etc,
and
it
basically
explodes
into
this
Explorer
widget.
D
That
allows
you
to
do
interactive
queries
and
set
like
visualize
different
types
of
visuals
and-
and
you
know,
set
up
those
visuals.
So
if
you
guys
have
seen
carrabelle
from
from
the
Airbnb
guys
that
have
like
a
nice
web
tool
for
doing
analysis,
it's
kind
of
in
the
vein
of
that,
where
you
basically
drop.
This
thing
pointed
to
some
data
and
you
can
visualize
it
and
query
it
to
some
certain
degree.
So
what
we're
trying
to
do
is
make
that
a
very
a
very
rich
widget
you
can
bring
in.
D
We
have
a
way
of
bringing
it
with
a
cleric
markup,
or
we
have
some
Python
Scala
and
our
bindings
that
you
just
run
a
function
called
Explorer,
give
it
a
data
frame
and
then
it
explodes
into
the
widget
now.
The
next
level
is
that
we
want
the
widget
to
be
to
be
able
to
tie
two
other
declarative,
widgets,
bye-bye
through
selection
and
and
sort
so
we're
going
to
enable
that
same
support.
So
you
can
basically
think
about
a
scenario
where
you
you
have
some
data
in
whatever
data
frame
flavor
are
pandas
spark.
D
Then
you
drop
this
explorer
widget.
You
do
interactive,
filtering
and
group
buys
with
with
the
UI
you
select
the
visualization,
and
then
you
can
actually
select
on
a
pie
wedge
or
on
a
bar
chart
and
tie
that
over
to
an
extra
island
of
declarative
with
the
tree
and
keep
going
along
doing
your
your
notebook.
So
I
am
gonna,
create
a
screencap
and
I'll
post
it
on
the
half
or
maybe
somewhere
in
the
readme
and
then
point
to
it.
So
you
guys
can
take
a
look
at
the
progress
we
made
so
far.
D
A
Cool,
so
I
actually
just
makes
me
think
of
something
that
is
kind
of
a
question
both
for
for
you
guys,
but
also
sort
of
her
for
Steve
and
Chris
and
darienne,
which
is
to
what
extent
is
it
now
easy
on
lab
to
actually
say
I
want
this
particular
widget
to
pop
out.
I
know
that
the
widgets
are
not
fully
working
yet
kind
of
the
existing
widgets,
there's
still
a
few
kinks
that
need
to
be
ironed,
but
but
to
what
extent
are
we
close
to
being
able
to
say?
A
I
want
one
of
these
things
actually
not
in
line
in
the
notebook
but
displayed
on
a
separate.
I
think
chris
use
the
word
property
the
other
day.
That
may
be
the
fix.
The
best
word
that
we
have
right
now
on
a
separate
thingy
right
on
a
separate
sub
tab
in
lab,
while
maintaining
communication
with
the
colonel,
so
that
taking
advantage
of
the
lab
real
estate
in
the
lab
you
I.
A
E
E
E
A
F
A
A
G
Thing
I
want
you
to
bring
up
is
that
there
will
be
a
temptation
to
create
new
plugins
of
viewers
like
a
lot
of
them
and
which,
because
I
mean
the
object
model
off
I
mean
chip
that
I
bellows
it.
But
if
we
get
diverted
into
this
sort
of
thing-
and
we
do
a
lot
of
this-
we
want
to
take
more
time
to
reach
feature
parity
in
the
notebook
itself
and
yeah.
It's
a
sealed.
Under
there
I'm.
G
So
that,
then,
anything
is
really
great.
I
mean
I'm
talking
about
the
viewers
like
the
geo,
Jason
viewer,
and
things
like
this
that
are
really
fancy
and
nice,
but
for
this
thought
of
graphical
users,
I
think
that
an
interaction
that
uses
that
also
interacts
with
the
colonel,
like
Jupiter
widgets,
is
much
more
valuable
than
just
a
viewer.
Liquefier
you
were.
That
is
completely
static.
G
A
Ya
know
for
the
the
things
that
we
want
to
hit
and
I
and
I
actually
want
to
now
that
we're
on
the
call.
I
want
to
encourage
people
to
be
fairly
disciplined
and
focused
about
tagging
as
090
milestone
the
things
that
are
really
key,
missing,
missing
features
or
bugs
or
problems
that
make
it
not
viable
to
use
to
use
lab
as
an
everyday
replacement,
and
that
would
get
in
the
way
of
a
regular
user.
A
And
so,
let's,
let's
use
the
090
tag
for
that
purpose,
so
that
we
try
to
stay
focused
on
this
on
on
that.
But
if
the
community
starts
building
these,
these
are
the
things
that's
great,
because
it
also
I
mean
we
want
to
get
adoption
from
the
broader
community
of
the
architecture,
and
so
but
but
we
should
remain
focused
on
getting
getting
on.
A
A
A
Yep
I
have
it
open,
so
my
only
notice
to
remind
folks
that
we
will
have
the
Jupiter
health
workshop
on
Friday
and
and
so
will
record
it
we're
going
to
have
it.
The
repo
that
I
linked
in
there
has
has
the
basic
info
Oh,
Jamie
and
I.
You
yesterday
get
the
tests
of
the
video
call
we'll
be
using
some
people
zoom
instead
of
blue
jeans,
same
difference,
which
is
the
one
that's
installed
by
default
on
the
big
conference
rooms
here
and
I'll.
A
I
You
know
so
there
are.
There
are
some
issues
coming
in
and
some
of
them
are
council
related,
so
I'm,
looking
at
those
in
particular
I,
think
there's
some
testing
I
need
to
do
in
IE
for
the
layout
of
completion,
widgets
and
tooltips,
and
that
affects
notebook
as
well,
obviously,
but
just
from
sort
of
anecdotal
evidence,
it
seems
people
are
displeased
with
tooltip,
which
I
kind
of
thought
they
would
be.
I
was
to
begin
with,
but
I
just
wanted
something
there.
I
I
You
right
well,
listen.
I
asked
I
asked
in
multiple
meetings
and
I
specifically
said
in
multiple
meanings
that
I
put
it
there
to
annoy
you,
but
no
one
really
engaged
me
with
it.
So
I'm
gonna
make
a
decision
and
maybe
they'll
annoy
people
less
and
if
it
does
we'll
stick
with
it,
if
not,
will
iterate
from
there.
But
I
I
was
willing
to
annoy
all
of
you,
but
now
that
there
are
thousands
of
users
I'm
less
willing
to
let
it
slide.
I
J
H
A
Python,
which
I
thought
but
I
would
actually
say
something
maybe
a
little
bit
different.
The
way
I've
been
thinking
about
the
a
lot
of
these
UI
decisions,
as
we
as
lab
begins
to
congeal.
Is
that,
rather
than
trying
to
sort
of
improve
a
lot
with
with
new
ideas,
our
first
order
of
business
is
to
really
get
as
quickly
as
possible
through
the
transition,
and
so
one
way
to
do
that
is
to
just
reuse
what
we
have
and
make
decisions
that
we
already
know.
A
At
least
people
are
used
to,
and
so,
for
example,
for
tooltips
I
agree
that
the
current
tool
tip
is
alone.
Knowing
what
we
could
do
is
simply
take.
The
decision
of
the
notebook
for
the
tool
type
is
something
that
you
pop
up
with
shift
tab,
and
since
now
the
console
is
closer
to
the
notebook
than
it
used
to
be.
A
We
could
just
follow
the
same
decisions
and
that
way,
we're
reusing
the
same
code
where
you
work,
it's
the
same
pattern
and-
and
we
can
get
there
with
basic
decisions
that
at
least
people
are
used
later
we
can,
we
can
think
about
how
to
make
them
different.
How
to
make
them
better,
etc,
but
that
way
we're
not
we
don't
have.
We
don't
have
to
iterate
as
much
right
now
on
decision
making
out.
We
can
just
go
with
what
we've
done
in
the
past.
Oh
because
I
don't.
B
A
That
I,
don't
I,
think
with
the
tool
tip,
is
one
area
of
where
I
there's
a
couple
examples
where
I
don't
think
we
should
implement
what
we
have
done.
One
is
collapsing
output.
That
UI
is
not
it's
really
problematic
and
difficult,
and
the
other
is
tooltips
and
I
like
I.
Think
it's
I
think
we
should
have
one
type
of
you
I
that
pops
up
when
users
are
typing
back.
J
A
J
C
A
C
H
J
I
soon,
as
soon
as
I
bite
on
five
min
during
the
beta
cycle,
we
got
our
sees
that
broke
the
treaty.
We
had
issues
on
I-59
repository
in
minutes,
but
but
I
paid
for
doesn't
have
to
all
trips.
No
but
I
pythons
I
python
api
broke
the
Tajiks.
The
two
teeth
would
basically
say:
object,
object,
internal
booty,
so
I
fight,
okay,
I.
J
C
I
I
have
a
great
forward
fix
and
nine,
and
if
that
straightforward
fix,
which
won't
take
a
lot
of
time,
is
not
to
our
liking,
we
can
disable
it,
but
I
I,
don't
think
we
need
to
bike
shed
forever
about
it.
It's
not
I'm,
not
I'm,
not
going
to
go
waste
a
bunch
of
time
on
it.
We
all
have
better
things
that
we
need
to
work
on
yeah.
A
Even
if
it
wasn't
perfect,
we
we
have
to
have
a
very
MVP
mindset
right
now
and
even
if
what
we
have
is
suboptimal,
people
are
using
it
and
we
harm
the
project
more
by
by
lengthening
this
gap,
then,
then,
by
giving
people
what
they
already
have
so
I
under
I
understand
your
concern.
Brian
about
about
there
are
things
were
what
we
have
is
really
saw
optimal,
but
we
need
to
be
careful
about
balancing
how
much
effort
we
put
into
rebuilding
that
right
now
and
I'm
not
saying
about
it.
E
I
need
to
jump
in
there
and
say
I
really
disagree
with
the
MVP
mindset
like
we
had
such
a
smash
hit
that
scifi
and
it
went
over
so
well
and
new
users
are
finding
it
so
easy
to
use
specifically
because
we
did
not
build
this
with
an
MVP
mindset
and
so
as
soon
as
we
start
taking
the
approach.
Well,
let's
just
hack
something
in
there
to
get
what
we
have
running
like
that
leads
to
inevitable
technical
debt,
which
will
hurt
us
in
the
long
run,
I'm
not
set.
E
What
I'm
not
saying
is
that
we
shouldn't
try
to
be
featured
parroting,
that
we
shouldn't
try
to
make
things
better
and
we
shouldn't
try
to
be
done
in
a
reasonable
amount
of
time.
But
I
would
disagree
with
approaching
the
problem
in
that
mindset,
because,
in
my
entire
career,
I've
never
seen
that
work
out.
Well,
no
long-term
I.
A
Think
the
other
part
that
really
is
important
to
me
is
the
only
way
to
answer
these
questions
is
to
do
user
tests.
Yes
and
like
I'm,
all
for
I.
Think
we've
done
that
by
having
the
existing
tool
tip
in
Jupiter
lab
and
people
have
been
frustrated
with
it,
let's
iterate
and
do
more
user
tests
and
I'm
fine.
If
that
iteration
is
either
closer
or
further
away
from
the
existing
implementation
and
the
classic
notebook,
but
I,
don't
think
we
should
somehow
a
shortcut
that
user
testing
and
implement
something
that
is
actually
confusing
to
users
that
that's.
A
A
All
of
those
I
am
in
favor
of
my
only
concern
is,
if
we
do
that,
and
we
have
the
lab
/
classic
split
for
two
years
from
now.
That's
a
problem
right
and
so
I'm.
Just
saying
that
right
now
we
are
in
a
sort
of
ground
splitting
between
our
legs
situation
and
there's
a
there's
a
limit
to
how
much
we
can.
We
can
handle
that
and
I
really
don't
want
to
forget
about
that
right,
yeah
and
that's
where
I
think
where
I'm
more
concerned
is
bending.
A
G
A
G
A
A
But
what's
that
I
think?
What's
that
I
think
we
can
spend
like
it's
I
pie?
Obviously
people
spent
time
on
you
know
markdown,
rendering
and
csv
stuff,
and
do
you
JSON
so
a
little
bit
of
time
into
or
hi
what
I
impact
new
features
that
the
existing
notebook
doesn't
have
that's
fine,
but
in
terms
of
significant
amounts
of
engineering,
I
think
we
do
need
to
focus
on
making
the
existing
set
of
features
designed
really
well
usable
bulletproof,
and
today
those.
E
High
impact
players
are
parents
to
get
people
to
move
like.
If
you
remember
the
problem,
like
the
biggest
problem,
I
found
three
transition
was
that
there
was
no
incentive
for
people
to
actually
move
move
over,
but
Fernando.
Your
shame
before
your
use
case
with
the
copy
and
pasting
cells
and
the
ease
with
which
we
can
extend
stuff
like
these
are
all
big
shining
examples
and
incentives
for
people
to
blower,
so
I
think
there's
a
healthy
balance
between
the.
F
Between
the
two
I
would
say,
not
just
the
incentive
to
move,
but
that's
an
easy
place
for
people
to
have
an
entry
point
into
contributing
to
the
project
and
that's
what
we
saw
with
people
in
this
friends
and
so
so
spending
a
little
bit
of
time,
exploring
at
least
plugins
that
are
designed
to
expect
the
exercise,
architecture
and
and
helping
people
come
on
board
I
think
benefits
of
project
in
the
long
term,
but
I
agreed.
You
don't
spend
all
your
time
there
I
mean.
Obviously
we
have.
G
J
K
A
We're
trying
to
curate
that
on
github
with
this
0.9
and
1.0
milestones,
or
it's
a
work
in
progress,
though
okay
and
and
that's
that's
what
I
wanted
to
reiterate
that
people
like
tag.
If
you
see
something
that
is
of
that
nature,
please
tag
it
as
090,
so
that
that
milestone
remains
really
that's.
How
working
list.
L
Steve
did
anything
that
you
wanted
to
bring
up.
I
just
want
to
see
if
there
are
any
objections
to
consolidating
Jupiter,
j/s
utils
into
Jupiter
Jaya
services,
it's
one
less
dependency
and
actually
and
the
process
of
doing
that,
I'm
looking
to
get
rid
of
pieces
that
were
in
jupiter
jsu
kills
and
make
them
dependencies.
So
it's
just
a
solid
cleanup
activity.
I
think
and
Jason
shake
and.
F
A
J
Sighs
I
was
very
great.
It
was
really
good
to
meet
with
you
guys.
I
wish
I
had
more
time
to
actually
write
a
script.
I
don't
afford
a
single
line.
It
was
really
nice
to
see
how
you
can
have
it
was
like
then,
we're
able
to
get
things
in
a
few
minutes
and
22
everybody
who
did
TRS
and
we
based
things
at
the
last
minute
for
Toronto.
We
had
a
couple
of
other
talks.
We
have
a
talk
from
mean
about
NBD.
How
to
defend
were
not
book,
women,
envied
I'm
yeah
and
it
down.
I'm.
J
Sorry,
young
players
just
came
back
from
a
promise
to
see
it
should
be
already
online.
If
you
want
to
do
see
how
to
use
it,
and
we
have
also
the
repository
which
is
really
you
can
just
control
it
and
it's
set
up
get
books
and
you
can
see
that
books
ever
with
a
nice
UI.
It's
not
perfectly
of
design
work,
but
we
would
be
happy
to
get
your
feedback
on
that,
and
this
is
talk
about
any
field
and
how
to
do.
J
Reproducible
workflow
with
notebook
was
more
than
food
I,
with
people
to
depend
on
each
step
of
the
stairs
and
also
ground
just
hyper
feet,
because
we
really
great
and
so
a
lot
of
interests
for
this
project
system
is
also
online.
I
think
both
of
these
talks
will
have
links
and
some
rentals
on
a
bet
you
already
saw
several
post
on
our
on
our
blog,
would
be
in
the
newsletter
most
likely
that
should
go
out
in
a
few
day.
J
I
plant
I,
don't
remember,
and
we
had
a
lot
of
users
that
the
user
testing,
as
you
might
have
guessed
from
other
people
talking
before
me,
and
they
got
to
be
three
t-shirt.
So
if
you
see
people
with
a
new
t-shirt
that
you
haven't
seen
before
and
Jupiter
in
big
on
it,
they
are
the
one
who
add
us
to
refine
the
UI,
and
so
you
can
just
say
thanks
to
them.
I
thought
about
it.
A
K
A
A
A
We
put
in
its
this
case.
This
case,
that's
done
by
the
mega
line.
Renderer
is
the
one
that
produces
that
dandy.
Exactly
we,
the
Vega
light,
render
grabs
it
from
the
canvas
object
and
then
just
injects
it
as
a
new
mind
type
in
that
output,
and
that's
something
also
that
be
to
plot
needs
that
bouquet
needs,
and
so
in
Jupiter
lab.
We
probably
need
to
think
about
a
way
for
one
output
to
of
a
given
line
type
to
add
an
additional
mind
type
to
that
output.
A
F
F
C
F
A
Right,
do
you
want
to
maybe
open
an
issue
to
track
this
question
and
it's
almost
it's
almost
kind
of
at
the
somewhere
between
the
protocol
level
and
the
implementation
right,
because
there
may
be,
there
may
be
some
messages
that
need
to
be
a
no,
no,
no
I
don't
do
it.
I,
don't
think
it's
protocol
level,
it's
more
just
as
sort
of
what
JavaScript
API
is
available
to
JavaScript
output
is
sort
of
the
question
to
update
the
output,
but
at
least
at
least,
let's
make
sure
that
we
document
this.
A
The
reason
why
I'm
thinking
of
the
protocol
is
because
the
protocol
kind
of
presents
the
complete
model,
and
up
until
now,
we've
described
a
compute
model.
Basically,
as
the
colonel
produces
an
output,
you
bundle
it
and
you
send
it
and
you're
done
right
and
that's
it
so,
once
you've
created
that
little
package
you've
sent
it
it's
in
an
envelope
and
that
envelope
went
went
and
it's
finished.
A
This
introduces
a
change
in
that
model
in
that
that
envelope
that
the
content
of
that
envelope
may
get
updated
client-side
after
the
fact,
and
the
fine
version
is
not
what
the
colonel
produced
so,
where
we
describe
the
description
of
the
core
kind
of
compute
model
and
representation
model
should
should
explain
this.
We
can
prototype
what
it,
how
it
should
work
client-side
until
we
get
something
that
works
well,
but
that
piece
needs
its
kind
of
at
the
very
bottom
of
the
architecture
yep.
A
What
happens
in
the
client?
What
ends
up
in
the
document?
Those
pieces
need
to
be
clearly
described.
Yep,
obviously,
there's
a
note
thinner
about
the
user
test.
Student
bit
of
students
did
an
absolutely
fantastic
job
and
organizing
user
tests
at
sci
fi.
They
actually
had
by
the
end
over
a
hundred
people
who
wanted
to
do
in
person.
J
Else
have
any
other
comments.
Turnouts
I
was
just
wondering.
If
you
have
hundreds
of
users,
can
we
ask
them
to
do
or
not
I
mean
to
do
user
testing
on
their
hand
and
say
hey
P
is
if
you
want
to
do
record
your
screen
right.
We
have
to
do,
and
here
is
this
clear:
the
script
can
you
can
you
send
it
to
us
and
we'll
try
to
improve?
You
can
talk
doing
that
yeah
so.
A
We're
the
plan
is
to
spend
some
time
addressing
the
usability
problems.
We
found
at
side
pie
and
do
another
round
of
user
tests
and
take
advantage
of
those
volunteers
who
were
willing
and
do
it
online.
Like
you
said
it
was
really
encouraging.
So
to
see
that
response
great
great
any
other
notes.
Folks,
remember
anyone.
A
So
last
but
not
least,
I
want
to
make
sure
that
we
introduce
a
new
a
new
face
for
the
the
public
meeting,
even
though
many
of
you
did
get
to
meet
the
meet
Mike.
Who
was
a
who's
here
with
us
today?
So
Mike
is
starting
in
the
fall
as
a
new
post.
Doc
he's
a
kind
of
hard
time
with
us
right
now,
while
he,
while
he
finishes
his
PhD
here
at
Berkeley,
and
it
will
be
very
soon
working
working
full
time
with
us.
A
I
hope
you
had
a
good
experience
at
sci
fi,
my
soul
will
have
a
chance
to
debrief
later
and
so.
I
want
I
want
folks
to
kind
of
welcome,
Mike,
13
and
so
they'll
be
with
us.
So
focusing
a
lot
on
the
more
documentation,
publishing
notebooks
as
literature,
part
of
the
project
and
kind
of
integrating
that
into
kind
of
the
the
education
and
academic
publishing
pipelines
question
is
that
we
had,
on
the
run.