►
From YouTube: Jupyter/IPython Dev Meeting, November 29, 2016
Description
Meeting of the Jupyter/IPython development team, November 29, 2016.
Meeting Notes: https://jupyter.hackpad.com/November-2016-weekly-meetings-ODhDM77jlsd
A
Good
morning,
everyone,
this
is
the
Tuesday
November
29th
after
the
u.s.
Thanksgiving
holiday,
ipython
Jupiter
that
meeting
and
I'm
sort
of
coming
back
from
the
dead
have
been
on
the
road
for
about
three
months.
Almost
in
continuous
travel,
so
I'll
be
mostly
catching
up
with
what
the
team
has
been
up
to
and
may
see.
Some
new
faces,
so
I'm
kind
of
excited
to
see
some
of
the
reports
and
then
it
work
and
I
know
that
there's
been
changes,
the
meaning
format.
A
C
This
was
probably
Thomas
but
I.
Think
Thomas
is
traveling
with
poor
internet
right
now.
So
Dylan
we're
just
a
couple
of
issues
away
from
a
notebook
4.3
release,
so
people
can
pitch
in
list
a
couple
remaining
issues
or
decides
to
to
punt
them
and
go
through
and
make
sure
that
we've
got
some
changelog
highlights
and
things
for
for.
What's
new,
then
we
should
have
a
notebook
release
in
a
couple
weeks.
D
So
there's
a
new
set
of
features
that
aren't
fully
baked
in
yet,
but
almost
there
that
have
to
do
with
restoring
the
state
of
Jupiter
lab
when
you
refresh
the
page,
so
the
P
R,
that's
open.
Now,
there's
one
outstanding
bug
that
I'm
trying
to
fix,
but
basically
what
it
does
is
when
you
refresh
the
page,
every
single
widget
that
you
had
opened
before
reinstates
itself
and
the
widget
that
had
the
focus
gets
the
focus.
D
D
We
have
the
remembering
which
one
was
current
and
we're
getting
a
bit
better
about
handling
if
it
can't
open
up
one
sort
of
just
forgetting
that
in
the
database
there's
a
there's
a
long
overview
of
the
actual
life
cycle,
steps
that
it
has
to
go
through.
That's
in
the
text
of
the
P
R
that
I
linked
in
the
hackpad,
but
the
upshot
for
most
plugin
authors
is
really
that
if
they
just
use
the
baked-in
instance
tracker,
all
of
the
state
handling
is
basically
free.
D
They
just
have
to
say
here's
how
you
generate
a
unique
name
for
the
type
of
widget
that
I
am
so
for
notebooks,
for
instance,
it
just
returns
a
path
and
they
have
to
say
here's
how
you
generate
the
arguments
that
the
command
that
creates
me
would
need
to
recreate
me
and
that's
pretty
much
it.
The
restoration
can
then
automatically
happen.
D
We
take
care
of
all
the
complexity
so
later
in
the
week
once
this
stuff
is
merged,
I'd
encourage
you
all
to
play
with
it
that
the
fact
that
it's
brand-new
and
it's
fairly
volatile
means
that
just
just
for
sanity's
sake
in
the
help
menu
and
in
the
command
palette.
There
are
commands
that,
let
you
clear
the
application
restore
state,
because
you
might
find
yourself
locked
in
some
sort
of
bad
state
and
you
just
need
to
get
rid
of
it.
Somehow,
especially
if
you
opened
up.
D
If
you
started
your
server
in
a
different
directory
one
time
than
you
did
another
time,
that
sort
of
thing
that
might
cause
bad
state,
so
you
could
use
that
help
menu,
just
go
I
made
it
the
last
item
in
the
help,
menu,
clear
application,
restore
state
and
it's
stored
in
your
local
storage.
So
if
you
tell
your
browser
to
forget
everything
or
you
open
a
new
incognito
window,
it
won't
remember
that,
but
at
some
point
we'll
probably
switch
it
to
some
different
storage
mechanism.
B
A
Reception
from
people
I
think
we're
going
to
have
potentially
some
very
interesting
collaborations
that
may
pop
up
over
the
next
few
months
that
may
sort
of
materialize
work
with
others
wanting
to
contribute
to
the
system
sort
of
large
scientific
projects
wanting
to
contribute
to
the
system.
Overall,
the
demos
worked
very
well
and
the
architecture
is
kind
of
what
people
need
at
the
I
spoke
at
the
gateways
conference,
which
is
a
kind
of
supercomputing
web
scientific
web
portals.
A
What
gateways
interfaces
for
science-
and
there
was
a
lot
of
interest
from
from
a
variety
of
centers
in
in
these
tools.
Matt
Rocklin
also
prepared
a
really
nice
task,
demo
that
uses
that
provides
tasks.
Visual
real-time
tasks,
visualization
with
Jupiter
lab
plugins.
That
works
extremely
well
and
kind
of
highlights
exactly
why
the
design
kind
of
is
the
way
it
is,
and
why
that's
a
good
thing
certain
use
cases
so
I
just
want
to
pass
that
on
to
the
team
that
it
really.
C
C
Tuturro
like
seven
and
NP
dime
is
in
a
similar
state
with
I
need
to
assemble
screenshots
and
Carols.
Something
out
on
some
text
in
in
the
docks
and
Martin
and
Veda
are
working
on
some
west
merge
conflict
resolution
things,
but
it's
in
pretty
good
shape
and
it
should
be
out
well.
It
will
be
out
this
week
because
we'll
be
talking
demoing
out
and
helping
people
with
it
in
France
at
the
chipper
de
Paris
next
week,
yep.
B
C
C
He's
gonna
actually
go
through
the
release
process
and
then
and
we're
gonna
help
and
see
if
there's
anything
that
any
information
missing
so
kind
of
formalizing
the
handing
off
of
that
project.
But
it's
it's
ready
to
go
we're
just
waiting
to
spend
a
little
time
to
actually
go
through
the
release.
Steps.
B
B
F
B
F
F
About
that
we're
just
working
really
hard
on
trying
to
get
a
6x
release
out
for
iPad
widgets,
doing
a
lot
of
backwards,
compatibility
and
styling
overhaul
and
someone's
working
on
the
embedding
work.
Just
trying
to
make
the
transition
from
the
five
scenarious
to
six
series
go
well
because
there's
quite
a
bit
of
code
overhauling
that
we
did
for
six
and
we'll
keep
plugging
away
on
it.
A
For
the
benefit
of
anyone
listening
or
who
watches
the
video,
after
do
you
foresee
any
significant
backwards,
incompatible
changes,
at
least
at
the
Python
level
API
or
at
the
JavaScript
API?
Absolutely.
E
Yeah
at
the
Python
the
world,
there
will
be
some
different
ideas
that
would
be
definitely
removed.
So
that's
the
main
thing
yeah
and
the
JavaScript
side.
It
come
it's
completely
new
if
you
want
to
make
a
gift
or
a
budget,
and
the
main
backward
incompatibility
is
on
the
subject
of
default
values
and
the
JavaScript
side.
If
you
want
to
have
embeddable
widgets
syntax,
there
will
be
a
little
bit
different.
That
transition
should
not
be
too
hard.
A
E
Old
api's
work
everything
works
for
the
old
notebook.
You
just
need
to
change
the
way
you
define
default
values
if
you
want
to
have
a
short
realization
of
the
widgets
for
embedding
widgets,
because
so
that
you
can
skip
default
values,
so
that
is
the
main
backward
incompatible
change.
If
you
want
to
make
a
do,
peter
lab,
widget
library,
you
need
to
write
a
new
set
of
extension,
which
is
completing
you
yeah.
G
B
H
If
I
remove
me,
yes,
okay,
great,
so
sorry,
the
reason
echo
so
nothing
really
new
with
github
automation,
it
was
like
we
had
Thanksgiving
and
everything
and
I
asked
something
to
github,
because
it's
used
a
bit
I
di.
What
I
want
to
know.
It's
all
going
to
be
changes.
I've
been
mostly
working
on
putting
the
ipython
mastering
a
patent
for
book
and
I,
better,
not
org,
and
responding
to
packed,
because
the
discussion
was
going
nowhere,
so
I
I
responded.
It
was
already
on
website
last
last
week
and
still
no
feedback
on
the
book
policy.
H
So
if
you
have
one,
if
you
have
comments,
please
give
your
comments
soon.
Otherwise
we
will
assume
it's
a
final
policy
and
working
on
Python
talk
and
tutorial
proposal.
The
Toyota
Toyota
puzzle
are
due
tomorrow,
so
I
want
to
try
to
send
the
current
one.
Today,
I
gave
a
link
if
you
have
any
feedback,
please
try
to
speak
up
before
end
of
day,
especially
since
Mike
Wright
was
with
whom
I'm
going
doing
some
tutorial.
If
nothing
change
is
in
Europe,
so
won't
be
able
to
give
feedback
certain
and
also
working
on
the
talk
proposal.
H
But
this
one
is
getting
late,
also
not
working
too
much
from
that,
because
the
deadline
is
later.
I
would
like
to
have
feedback
and
work
on
it,
but
if
there
is
not
much
iteration
on
it,
I
will
just
write
it
myself
and
submit
it
and
that's
about
it.
Otherwise,
I've
been
working
on
grant
report
and
thing
like
that.
A
H
I'm
going
to
get
that
one
of
the
question
is:
who
can
go
or
who
would
go
to
Python
and
then,
depending
on
who
can
go?
We
can
decide
which
kind
of
subject
we
can
talk
about
because
otherwise
it's
a
vicious
circle.
You
don't
make
a
proposal
because
you
don't
know
who
is
going
and
who's
going
because
you
don't
make
it
over.
B
I
B
I
Thank
you
all
for
entertaining
my
little
input
here,
so
we
at
Mozilla
really
rely
on
Jupiter
notebooks
for
distributed
data
processing.
So
thank
you
to
everyone.
The
community
has
made
made
the
awesome
tool
available
and
we're
interested
in
seeing
how
the
notebook
might
be
modified
to
better
support,
distributed.
Data
processing
and
we've
met
with
a
couple
folks
from
the
project
about
that
and
into
another
meeting
or
two,
and
because
the
notebook
is
so
enormous,
ly
popular
and
because
there
are
so
many
different
groups
using
the
tool.
I
In
this
way,
we
thought
it
might
make
sense
to
try
and
get
some
evidence
for
how
the
tool
is
used.
So
we
wrote
a
Python
script
that
will
take
a
collection
of
notebooks
and
extract
features
from
it
and
allow
you
to
do
data
analysis
on
how
notebooks
are
being
used.
So
it's
sort
of
metadata
analysis
or
data
fooding.
Your
idea
or
whatever
metaphor
you
want
to
use
and
what
we
have
now
is
a
very
simple
proof
of
concept
written
by
conner
Amyris,
who
was
not
able
to
be
here
today.
I
A
Just
have
a
question
this
is
this
is
great
and
I'm
very
excited
to
see
it.
Sam
thanks
for
pitching
in
quick
question,
are
you?
How
are
you
do
you
have
a
classifier
that
you're
training
and,
if
so,
is
it
supervised
with,
with,
with
with
labels
that
you
provided
in
advance
on
or
how
are
you
doing,
the
classification
right.
I
Now
it's
a
very
crude
k-means
cluster,
but
this
is
really
proof
of
concept
level
code
like
there's
nothing
magic
to
it.
We
go
through
and
we
set
a
boolean.
Did
it
read
data
from
s3?
Does
it
appear
to
use
PI
spark
which
spark
is
the
distributed
processing
framework
and
that
gets
that
the
1.0
if
it
does
and
it
gets
into
0.0
if
it
doesn't,
and
it's
there
about
10
features
I
think
in
our
example
data
set
and
then
we
just
run
a
very
crude
clustering
algorithm.
I
So,
though,
the
value
isn't
in
the
specific
clustering
we've
done,
the
potential
value
is
have
we
shown
an
approach
that
other
people
can
adopt
and
then
start
to
do
the
usual
dance
of
picks
and
features
run.
Some
analysis
realize
that
your
features
should
have
been
different
run.
Some
clustering
realize
that
your
model
approaches
they
different
all
that
stuff
that
people
do
all
day.
J
That's
fantastic:
this
is
super
in
line
with
some
of
the
things
that
I've
been
talking
about,
Emma
at
the
morsel
of
stuff
and
also
with
other
people
and
UV
and
I
had
been,
and
Stewart
and
Mattias
had
started
some
path
down
getting,
rather
than
dealing
with
the
analysis.
First,
we
were
going
to
try
to
scrape
every
notebook
off
of
github
that
is
publicly
accessible
and
would
seem
to
be
once
that
dataset
is
assembled
would
be
in
line
with
this
kind
of
we
were
doing
it
with
this
sort
of
work
in
mind.
J
I
You're
so
welcome
and
yeah
Connors
done
a
little
bit
of
github
scraping
also
and
I
bang
my
head
against
it.
There
might
be
some
room
for
for
us
to
sort
of
organize
and
collaborate
on
data
collection,
since
that
can
be
a
big
pain
in
the
butt
and
another
thing
is
getting
organized
on
outputs.
You
know,
hey
I
went-
and
you
know,
looked
at
this
collection
of
800
desk
notebooks
and
here's.
What
I
found-
and
here
was
the
script
I
used
blah
blah
blah
blah
blah.
I
J
I
J
So
the
thing
that
is
most
relevant
to
share
is
some
further
movement
forward
on
header
ID
filtering,
specifically
trying
to
get
it
to
parity
or
mostly
parity
with
pan
doc
and
figuring
out.
What
that
actually
means
is
fun,
and
so
in
Python.
This
is
going
to
require
using
the
regex
library,
not
re,
because
we
need
access
to
unicode
property
classes.
I'm
trying
right
now
to
find
a
good
jos
library
for
this
x-ray
ax
deems
like
the
best
thing
that
we've
got
going
for
us,
but
I
could
be
I,
don't
actually
know
this
subfield
as
well.
J
B
K
L
Have
nothing
major
to
report
fourth
quarters
very
busy
around
here,
so
I
haven't
had
a
lot
of
cycles
and
spent
hundreds
but
yeah
I'm,
hoping
me
and
I
know
you
got
back
to
me
on
the
MV
format.
Er
and
I.
Just
wrote
back
to
you.
I
will
go
and
make
those
updates
and
I
think
I
opened
one
Jupiter
hub
of
about
a
user
name
field
and
if
I
find
some
cycles,
I'll
try
to
jump
on
that
man.
That's
about
it!
For
me,.
A
I'm,
getting
a
hammering
or
from
Jupiter
I
guess:
I
guess
the
highlight
of
it
is
that
as
I
said
not
just
about
to
provide,
but
that
in
general
the
the
response
around
various
venues
continues
to
be
very
positive
and
I
I
I
try
to
tell
people
that
everything
that
I'm
saying
the
credit
goes
to
all
of
you
guys
and
not
to
me,
but
I
want
to
reiterate
that
here
with
the
team
so
but
yeah.
Otherwise,
it's
been
mostly
granulated
stuff.
A
We
have
the
project
report
coming
due
this
week,
which
I
really
appreciate
the
help
that
you
folks
all
gave
me
through
the
dead
meeting
and
I
know.
A
number
of
you
are
still
working
on
that
because
it's
making
the
process
a
lot
easier
and
other
various
kind
of
project
admin
related
things,
but
no
it's
good
to
be
back
in
my
travel.
My
travel
frenzy
is
over
so
I'll,
probably
much
more
regularly
around
thanks.
B
H
One
request:
can
we
try
to
clean
up
the
notes
and
I
can
take
care
of
part
of
that
and
try
to
send
that
to
the
waiting
list,
because
I
think
that
having
a
nice
return,
salary
which
is
not
on
the
hackpad,
would
be
nice
to
engage
community
because
watching
the
video
to
catch
up
with
everything
take
I
mean
here.
It
would
take
close
to
half
an
hour
while
written
notes
are
faster
to
go
through.
H
H
A
Do
you
want
to
take
a
cut
at
that
and
see
see
sort
of
how
much
effort
it
entails
because
I
mean
the
problem
with
these
things
is
obviously
that
if
we
add,
if
we
add
after
every
meeting
a
task
that
ends
up
being
half
a
day,
it's
not
gonna
get
done.
So,
if
you
want
as
an
experiment,
why
don't
you
give
it
a
try
and
see
how
like
how
much
work?
A
Is
it
to
do
that
and
and
if
we
can
manage
to
do
it
in
a
very
lightweight
way,
I
think
it's
a
very
good
idea,
but
but,
as
always,
people
are
pretty
cycle
constrained
in
general
and
so
the
balance
of
I
just
I'm
being
sensitive
to
the
fact
that
it's
hard
to
commit
people
to
even
extra
work.
But
if
you
want
to
do
an
experiment
and
let
us
know
sort
of
getting
something
readable
takes
15
minutes,
that's
one
thing
getting
something.
H
Yeah
I
mean
I'm
also
doing
that
now,
so
that
for
the
next
meeting,
if
you
can
pay
slightly
more
attention,
when
you
take
notes
that
it
can
be
easier
to
turn
into
text,
it
would
be
easier
I
doubt
I
can
do
it
perfectly
this
time,
but
I'm
feeling
like
a
small
effort
on
a
couple
of
weeks.
It
would
be
nice
but
I
will
give
it
a
shot
and.
A
If
we
can
try
to
keep
them,
keep
the
notes
a
little
bit
more
readable
in
that
way,
as
we
go,
that
that
certainly
could
help
so
or
maybe
an
alternate
way
to
say
it
is
see
what
see,
what
can
you
do
in
30
minutes
and
put
a
clock,
see
what
you
can
do
in
30
minutes
and
and
we'll
see,
we'll
see
what
we
can,
what
we
can
get
done
if
we
cap
it.
If
we
cap
this
task
with
a
with
a
pretty
tight
time
limit,
how
does
that
sound?
How
does
that.
B
Some
periodic
like
tutorial
sessions
for
the
team,
they're
completely
open,
so
one
on
jupiter,
la
jupiter,
love,
jupiter,
lab
jupiter
hub
and
then
also
project
management
of
life,
and
that
one
you
know
I
would
hope
to
bring
together
folks,
like
Mateus
and
Jason
and
Carol,
and
all
the
people
who
have
provided
really
great
suggestions
on
things
that
we
could
implement
to
help
communicate
progress
across
all
the
different
profit
sub
projects.
So
we'd
love
to
do
that
in
the
next
couple
weeks.
B
A
A
That
could
be
probably
relatively
low
effort
for
folks,
like
Carol
and
min
and
UV,
who
kind
of
know,
know,
know
the
ins
and
outs
for
the
system
very
well,
but
it
would
be
very
valuable
to
to
outsiders
would
be,
would
be
really
good
thing
to
have
and
I
think
if
we
get
into
a
habit
of
doing
these
things
online
lightweight
relatively
informal.
So
it's
not
a
huge
burden
for
us,
but
but
we
do
them
in
a
way
that
really
helps
the
community
I.