►
From YouTube: Jupyter/IPython Dev Meeting, September 13, 2016
Description
Meeting of the Jupyter/IPython development team, September 13, 2016
Meeting Notes: https://jupyter.hackpad.com/September-2016-ncdE6E6XQmc
A
Good
morning
this
is
the
september
13th
jupiter
dev
meeting
we've
got
a
great
lineup
of
folks
today,
so
I
guess
we'll
start
with.
Should
we
start
with?
Let's
start
with
introducing
one
of
our
newest
team
members
who
just
joined
bloomberg
jason,
would
you
like
to
do
the
introductions.
C
Hey
so
I
went
on
a
really
long
bike
ride
for
the
last
two
years.
I
guess
the
project
name
has
changed,
so
that's
cool
paul
even
off.
I
also
go
by
pie.
Yeah
just
joined
bloomberg,
it's
my
second
day.
So
there
you
go
so
I'm
back
watch
out.
C
C
Sure
so
I
used
to
be
a
python
core
contributor.
I
guess
I'm
still
an
ipad
core
contributor,
but
I
used
to
work
on
the
notebook
used
to
work
on
ipython
core
used
to
contribute
to
matplotlib
dropped
off
doing
all
those
things
for
the
last
two
years.
While
I
was
working
at
discuss,
but
now
that
comments
on
the
internet
are
going
on
their
own,
I'm
back
to
work
on
jupiter
again.
D
Yeah
just
a
couple
releases
of
things,
traitlets
and
core
in
the
last
few
days.
The
main
thing
that
I'm
going
to
be
working
on
that
it
will
be
useful
from
people
in
general
is
the
new
stuff
for
the
next
release
of
jupiter
hub
is
now
all
merged.
So
I'm
just
working
on
testing
that
and
testing
with
the
np
creator
and
all
the
bonners
and
things
and
docs
for
updating
and
everything,
and
that's
the
main
thing
I'm
doing
for
the
next
little
while.
E
Hey
nothing
serious,
just
saying
that
python36
beta
1
is
out
if
we
can
test
with
multiple
package
to
make
sure
that
all
the
features
we
use
are
available
in
36
and
there
is
no
regression
it's
a
time
to
complain
so
that
things
that
have
been
removed
from
python
or
added
don't
break
any
of
our
code.
E
Typically,
these
have
been
some
syntax
warnings
that
now
are
syntax
errors
that
broke
some
packages,
and
we
also
have
some
pretty
nice
new
features
only
on
360.
That
would
be
really
nice
for
users
like
now,
you
have
a
class
of
import
error
which
is
module
not
found,
so
we
might
be
able
to
detect
when
the
user
has
not
something
installed
or
when
something
failed
to
import
anyway.
E
That's
about
it
for
me
just
if
we
can
make
sure
on
travis
that
to
test
on
on
36,
once
it's
available
and
and
complain
upstream
as
soon
as
possible.
It's.
E
It's
normally
feature
freeze,
so
we
don't,
they
won't
add
any
features,
but
if
something
is
basically
relied
upon
on
too
many
packages
they
might
roll
that
back.
In
particular,
one
of
the
latest
change
is
compact
ticked,
like
dictionary
now
in
c,
python
are
ordered
to
preserve
the
insertion
order
as
a
side
effect
which
might
have
weird
effects
if
we
rely
on
a
key
not
being
ordered.
E
Actually
it
might
be
the
opposite.
Now,
if
you
ride
on
36,
it
might
fail
on
other
version,
but
there
have
already
been
a
few
segfaults
and
people
that
complain
about
that.
It's
a
pretty
deep
change.
B
F
Sure
we
created
a
new
repository
for
creating
third-party
extensions
and
also
the
the
way
they're
consumed
from
jupiter
lab
itself,
we're
using
that
from
jupiter
lab
to
build
the
core
plug-ins,
and
we
also
added
a
lab
extension
that
follows
the
same
sort
of
workflow
as
the
jupiter
notebook,
mb
extension
command
line,
script
so
and
then
jason
used
that
to
create
the
ipi
widgets
lab
extension,
and
he
can
follow
up
on
that.
But
basically
what
we
have
is
we're
for
most
third-party
extension
authors.
F
It's
they
just
run
a
script
with
their
entry
point
that
returns
a
list
of
an
array
of
plugins
and.
B
It
wasn't
me
it
was
him
okay,
so
I
guess
this
big.
Oh
there
he
is
okay,
go
ahead!
Yeah
my
network's
going.
B
F
They'll
just
run
our
script
and
our
use
of
webpack
is
an
implementation
detail,
but
for
more
advanced
use
cases
they
can
add
extra
webpack
config
for
their
for
their
specific
use
case,
and
that's
all
I'm
going
to
say
before
I
lose
my
network
again.
B
B
I
think
the
biggest
thing
for
us
is
an
announcement
that
we've
got
an
initial
system
in
place
after
several
iterations
and
hitting
roadblocks
and
solving
them
thanks
again
to
chris
and
brian
and
other
people
who
had
tons
of
helpful
feedback
and
for
for
design
and
architecture
and
stuff.
B
G
B
Yeah
yeah
yeah,
well
I'll,
put
we'll
put
a
link
into
the
to
the
pr,
so
there's
there's
a
pr
to
jupyter
lab,
which
makes
a
new
jupyter
lab
extension
command
and
then
there's
an
entirely
separate
repo
which
bundles
up
all
this
webpack
plugin
and
magic
and
everything.
B
And
then
and
then
I
can
put
a
link
to
the
to
the
working
ipi
widget
extension,
an
example
of
like
how
you
do
it
like
the
one
command
you
run
essentially:
cool.
Okay,
thanks
yep.
A
All
right,
let's
see
here,
jason,
did
you
have
anything
else.
You
wanted
to
add.
H
I'm
sure
I
can
introduce
myself
again
well,
I
mean
I
guess
in
in
in
the
context
of
this
I've
I've
been
involved
with
condo,
forge
for
a
little
while
now
and
and
am
comfortable
enough
with
connor
ford
that
I
feel
confident
volunteering
to
help
with
kind
of
packaging
for
any
and
all
needs.
If
you
guys.
B
I
have
tried
to
update
the
jupiter
lab
package
for
condo
forge
and
it
complained
it
had
a
circle,
ci
error
and
somebody
suggested
that
we
do
contact
smithy
on
it
again
and
I'm
interesting
problems,
because
I
need
to
generate
an
anaconda
token
or
something
like
that
anyway,
and
now
I'm
trying
to
wait
a
password
reset
request
or
something.
H
It's
it's
close
enough
that
I
know
exactly
what
you're
talking
about
okay
and
I'm
not
actually
sure
how
to
say
it
in
a
better
way.
So
it's
good.
B
A
H
So
I'll
I'll,
happily
give
it
a
shot,
and
anybody
else
can
jump
in
and
correct
me
where
I
will
surely
go
wrong
here,
but
essentially
it
got
started
to
make
condo
packaging
more
transparent,
a
lot
of
the
recipes
or
a
lot
of
the
condo
packages
that
were
distributed
by
continuum.
H
H
To
to
the
second
point,
I
guess
would
be
to
make
condo
packaging
more
open
and
and
transparent,
and
the
third
issue
was
that
there
was
a
repo
on
one
of
continuum's
deposits
or
on
one
of
continuum's
orgs
that
was
kind
of
a
dumping
ground
for
recipes,
but
they
weren't
necessarily
being
used
to
do
anything.
They
were
just
kind
of
there,
but
they
weren't,
mirrored
or
built
onto
anaconda.org,
and
so
condo
forge
is
attempting
to
fill
all
of
those
voids
by
having
a
place
where
anybody
can
submit
a
recipe
to
build.
H
It'll
go
and
build
on
circle,
ci
to
build
for
linux,
it'll,
build
on
travis
for
os
10
and
at
fair
for
windows,
and
those
three
services
are
used
to
basically
test
your
test,
your
build
and
make
sure
that
it
will
that
it
will
work
and
then,
after
it's
been
reviewed
and
merged,
you
get
your
own
feedstock,
which
is
essentially
a
place
where
you
can
control
who
has
commit
rights
to
the
recipe
and
time
the
recipe
is
updated.
New
builds
are
created
and
uploaded
to
the
condo
forge
channel
on
anaconda.org.
H
H
Every
bi-weekly
dev
call
that
is
fairly
involved
and
there's
a
couple
more
people
from
continuum
that
are
on
those
dev
calls
it's
it's
it's
not
like
an
official
continuum
project,
but
it
is,
it
does
seem
to
have
the
blessing
of.
D
A
Thanks:
hey:
how
about
we
hear
from
kyle
kelly?
Are
you
ready,
kyle.
H
Yeah
I,
like
literally
just
started
typing
some
notes.
I
noticed
I
could
have
shared
this
at
the
last
one,
but
I
missed
it
made
a
pr
for
mb
format.
So
now
we
can
do
embedded
raw
json
just
like
we
can
do
for
application
json.
But
if
you
have
a
custom
mime
type
that
is,
has
the
plus
json
suffix.
H
It
can
be
raw
in
the
notebook
format
instead
of
a
double
encoded
string,
so
the
ones
that
are
already
so
we're
already
using
these
within
interact
in
with
geojson
plotly.
And
then
I
have
a
prototype
vega
like
renderer,
so
I'm
guessing
it
would
work
much
the
same
for
like
the
for
jupiter
lab
for
the
way
the
extensions
are
built
for
the
mime
types,
so
yep.
B
Yeah,
the
the
notebook
format
wait.
So
this
is
this:
is
the
nb
format
spec
right
that
you're
that
you're
yep?
Okay,
so
that'll
involve?
Well,
I
mean
we
don't
want
to
do
anything
until
we
actually
release
the
environment,
because
that
involves
a
version
bump
and
backwards
compatibility
and
everything
right.
Man.
D
Yeah,
well,
we
we
already
in
the
in
nb
format.
We
already
said
that
json
stuff
shouldn't
be
double
encoded,
app
and.
D
Yeah,
and
essentially
all
this
is,
is
saying
for
custom,
but
you
could
say
that
this
is
relaxing
the
a
regular
expression
to
what
it
should
have
been
yeah.
I
agree
so
there's
kind
of
some
semantics
about
what
this,
what
that
actually
means
for
versioning
in
terms
of
the
spec,
as
opposed
to
the
python
implementation
of
it.
D
I
don't
know
we
should
probably
do
a
release
pretty
soon
and-
and
you
can
look
at
ad
implementing-
I
mean
you:
can
you
could
do
a
backward
income,
a
backward
compatible
version
that
always
works
just
by
checking
if
it's
a
string
yeah
so
that
that's
how
we're.
D
Describe
it
as
things
things
that
are
application,
whatever
plus
json
are
erroneously
double
encoded
in
the
current
implementation
that
that
really,
the
only
thing
is
in
the
json
schema,
it
doesn't
correctly
capture
json
content
when
validating
whether
it
should
be
a
string
or
not.
So
I
think
you
can
go
ahead
and
like
it.
Now
is
a
fine
time.
H
There's
a
little
bit
of
trickiness
there
too,
which
is
that?
Because,
because
the
previous
spec
was
a
multi-line
string,
which
means
there
could
be
an
array
of
strings,
that's
like
topical
really,
it
can
only
be
an
object
right
now
it
it
can't
because
it
couldn't.
H
Even
though
it
should
be
able
to
be
an
array,
okay,
it's
a
multi-line
string,
yeah,
exactly
right.
It
can't
be
yeah
okay,
so
I
I
think
I
noted
that
I
noted
that
in
the
pr
and
the
issue-
and
I
was
like
well
that
sucks-
but
that's
okay-
top
level
object's
still
fine
yeah,
but
in
the
future
we
should
probably
allow
it
to
be
any
json.
D
D
H
B
B
A
So
so
I
went
back
last
week
to
fix
it.
I
can
do
the
same
thing
this
week.
Yeah,
let's
see
I
wanted
to
just
say,
hi
thomas
welcome
back
from
your
travels
great
to
see,
you
he's
got
some
great
pictures
in
the
hack
pad,
hey
anything
you
want
to
add
today.
A
A
We
try
to
aim
for
once
a
week
just
to
go
over
general
project
stuff
and
out
of
the
meeting
that
we
had
last
week,
looks
like
we're
working
on
updating
or
writing
a
new
grant
report.
So
this
is
an
annual
grant
report.
That's
due
at
the
end
of
november,
so
we'll
be
working
on
that
and
reaching
out
to
people
for
feedback
we'd
like
to
do
it
in
a
google
doc.
A
Let's
see
here,
the
november
dev
meeting
is
we're
working
on
scheduling
that
in
the
last
our
last
jupiter
meeting
on
tuesday
last
tuesday,
we
mentioned
to
the
group
that
we
had
scheduled
this
during
election
week
in
the
u.s,
which
you
know.
This
is
an
important
week
for
us
here,
for
so
so
for
those
of
you
who
who
want
to
make
sure
that
you
get
your
vote
in
we're
presenting
three
options.
A
Event,
so
just
let
us
know
we'll
on
and
I
will
send
out
an
email
sometime
this
week
just
to
get
individual
feedback
if
you're
not
comfortable,
sharing
here,
grant's
working
through
triaging
issues
and,
of
course,
as
he's
doing
that
he's
seeing
a
lot
of
opportunities
for
streamlining
the
process.
A
So
you
may
see
some
updates
and
github
from
from
us
on
that
that
side
and
ian
rose
should
be
starting
any
day
now
so
he's
waiting
for
his
appointment
with
hr
here
at
berkeley,
so
you
may
see
him
on
the
next
tuesday
call
or
maybe
the
week
after.
I
I
am
on
the
call,
I'm
just
pulling
together
an
email
to
follow
up
on
your
comment
on
election
day
and
also
to
ask
people
for
any
recommendations
for
venues
in
new
york
that
they
may
have.
We
are.
We
do
have
some
options,
nothing,
no
winners
yet.
I
So
this
is
kind
of
the
last
week
that
we
have
to
add
any
additional
options
that
people
may
have
so
I'll
send
out
a
message
now-
and
I
think
fernando's
request
was
that
everyone
respond
in
the
positive
in
that
they
were
okay
with
the
meeting
dates
as
opposed
to
having
people
of
flag
issues.
So
I
am
going
to
request
that
everybody
respond
to
the
email
with
an
affirmative
and
that's
about
it
all
right.
A
Does
anybody
else
have
so
anyone
else
have
something
to
add
before
we
sign
off.