►
Description
Presented by Felipe Hoffa, Google.
“Data gives us insights into how people build software, and the activities of open source communities on GitHub represent one of the richest datasets ever created of people working together at scale.”—GitHub Universe 2016 With Google BigQuery anyone can easily analyze the more than five years of GitHub metadata and 42+ terabytes of open source code. Felipe Hoffa explains how to leverage this data to understand the community and code related to any language or project. Relevant for open source creators, users, and choosers, this is data that you can leverage to make better choices.
About GitHub Universe:
GitHub Universe is a two-day conference dedicated to the creativity and curiosity of the largest software community in the world. Sessions cover topics from team culture to open source software across industries and technologies.
For more information on GitHub Universe, check the website:
https://githubuniverse.com
A
Hello,
everyone
thank
you
for
joining
me.
I'm
Felipe,
Hoffa
I
be
living
in
San
Francisco
for
Villa
60
years,
but
originally
from
Chile
I
started
at
Google
six
years
ago
for
software
engineer,
but
for
the
last
40
years,
I
have
been
a
developer
advocate
for
those
of
you
who
don't
know
what
a
developer
advocate
is
a
developer
advocate.
A
Is
a
software
engineer
with
a
license
to
speak
so
I'm
here
to
talk
to
all
of
you
today
we're
going
to
analyze,
750
billion
events,
even
more
46
terabytes
of
code,
and
hopefully
we
will
have
a
lot
of
fun
with
it.
Follow
me
on
Twitter
find
me
if
you
have
any
questions
and
let's
get
started.
What
is
this?
What
do
you
see
here?
A
Any
word
code:
tensorflow
live
science,
there's
a
lot
that
we
see
here
with
your
life
science.
We
see
a
copyright
with
the
import
and,
if
we
zoom
out,
we
find
more
things
like
how
much
code
do
we
have
here
how
many
contributors,
the
number
of
pull
requests,
the
number
of
stars.
So
what
I
see
here
other
than
cold
I
see
data
a
lot
of
data,
big
data
so
and
we
want
to
analyze
it
and
who
wants
to
analyze
github.
Hopefully
you
that's
why
you
came
to
this
talk.
This
is
important
prop
for
project
maintainers.
A
If
you
own
a
project,
you
want
to
know
how
popular
it
is.
Who
is
following
you
how
to
manage
change?
How
to
have
new
8
API
see
if
your
project
is
healthy?
If
you
are
a
project
user,
you
may
want
to
know
their
same
things
or
what
other
related
projects
you
can
find
even
before
becoming
a
project
user.
You
are
a
project
chooser
when
you
are
choosing
a
project
between
all
those
alternatives.
A
You
want
to
find
what's
the
best
one:
what's
the
healthiest
one,
what
which
one
is
closing
issues
in
a
more
responsive
way,
etc
and,
of
course,
data
lovers.
If
you
love
data,
I
have
a
lot
of
data
here
that
you
can
analyze
research,
a
Feinberg's,
discover,
fix
the
world
etc,
and
we're
going
to
work
with
three
main
data
sets
that
you
might
know
already.
A
One
is
github
archive.
This
one
has
been
collecting
all
get
cab
events
for
several
years
and
you
can
find
a
huge
log
every
hour
of
everything
that
everyone
is
doing
on
github.
We
also
have
gh
torrent,
that's
a
similar
project,
but
it
goes
beyond
it.
Also
searches
collects
a
lot
of
metadata
around
github
projects,
for
example.
If
you
want
to
find
the
profile
of
each
user,
where
do
they
come
from
etc
and
they
publish
events
in
real
time?
A
If
you
would
like
to
do
analysis
in
real
time
and
then
we'll
learn,
repository,
we're
going
to
look
at
is
a
repository
of
all
of
the
open
source
code
on
github.
That's
notable.
We
are
moving
it
into
a
new
table,
so
we
can
analyze
it
very
very
quickly
and
how
all
of
these
got
started
back
in
2012,
my
colleague,
India
Gregorek
started
collecting
all
of
these
events
and
at
first
it's
pretty
easy.
Basically,
he
made
all
of
these
files
available
on
the
web.
A
You
want
one
hour
later,
you
download
one
file,
you
want,
you
want
a
full
day,
you
download,
24
files,
etc.
So
if
you
want
only
one
hour,
that's
pretty
easy.
You
can
get
it
in
less
than
a
second
in
0.1
seconds,
five
megabyte
of
data.
You
can
uncompress
it,
and
that
is
pretty
easy.
You
just
download
data
and
compress
it
on
your
computer,
but
that's
if
you
want
one
hour.
If
you
want
seven
years
of
data
at
least
rate,
then
you
are
dealing
with
more
than
two
terabyte
of
data.
A
A
So
at
that
time,
at
Google
we
were
releasing
this
project,
big
query,
who
knows
bigquery
excellent,
so
this
side
of
the
query
that
side
half
of
you
know
it
perfect,
I'm
glad
that
you're,
a
lot
of
you
know
it
for
you
that
for
those
who
doesn't
you
should
know
that
bigquery
we
will
allow
you
to
analyze
terabytes
in
seconds.
It's
a
cloud.
It's
a
data
warehouse
in
the
cloud.
You
don't
need
to
turn
on
anything.
It's
always
on.
It
will
scale
from
bytes
to
petabytes.
A
You
can
work
with
tableau,
our
Python
data
studio
power
bi.
Whatever
your
favorite
tools
are,
they
can
communicate
with
bigquery.
One
of
my
two
of
my
favorite
features.
You
can
share
data
with
anyone.
You
can
keep
data
private
yourself.
You
can
share
it
with
some
of
your
teammates
or
you
can
share
it
with
the
world
and
everyone
has
a
free
terabyte
every
month
to
run
queries
so
anything
that
I'm
going
to
do
to
the
estate.
You
can
go
and
reproduce
it
in
the
next
hour.
A
If
you
challenge
me
to
yes
turn
on
your
computer
log
into
bigquery,
create
your
account
and
you
will
be
able
to
run
queries
so
before
we
get
started.
Let's
look,
and
there
start
we.
This
is
github
universe.
You
might
know
what
stars
I'm
referring
to
what
are
the
projects
that
got
the
most
github
stars,
for
example,
last
year,
if
we
log
into
bigquery
I,
have
my
screen
loaded
here:
bigquery
can
communicate
be
arrested.
Also,
has
this
pretty
useful
web
UI
that
you
can
just
come
load
and
start
running
your
queries
here?
A
I
have
archive
loaded
per
day
per
month
per
year.
If
I
look
at
all
of
last
year,
I
will
find
a
table
with
760
gigabytes
of
data,
320
million
events
we
have
in
github
last
year
and
if
I
want
to
count,
then
the
stars
per
project.
I
write
a
sequel
query
like
this,
for
this
ripple
give
me
the
number
of
stars
during
2016
group,
the
first
column
or
thereby
the
second
one
descend
in
order
leave
me
20.
A
These
were
the
top
20
projects
by
number
of
stars
recorded
while
we're
event.
Oh
no
I
was
missing
something.
There
is
a
column
called
type,
because
I
only
want
to
watch
look
at
the
events
that
are
star
event.
That's
what
event
on
github,
so
the
top
project
last
year
by
number
of
stars
was
free
code.
Camp
I
hundred
eighty
thousand
stars,
followed
by
Google
interview,
diversity,
programming,
books.
Those
three
projects
are
learn:
how
learning
how
to
program
kind
of
project
how
to
interview,
and
then
you
have
others
like
Butte
and
so
flow
etc.
A
So
what
we
did
here
is
we
just
analyze
a
huge
300
million
events.
We
look
for
the
stars,
we
counted
them
and
we
got
our
ranking,
but
we
have
to
wait
a
minute
here.
We
have
to
look
at
what
we
are
doing.
This
is
not
the
number
of
stars
that
each
project
caught
last
year
because
if
I
go
to
hit
hub
here,
we
have
the
dataflow
project.
I
can
start
this
project
many
times
and
every
time
I
start
this
project
I'm
having
a
new
log
event
of
Felipe
Hoffa
start
this
project.
A
So
the
first
thing
I
have
to
do
when
running.
My
rankings
is
think
about
how
people
could
be
gaming
them
and
instead
of
counting
the
number
of
stars,
I
could
count
the
Distilled
actors,
how
many
people
have
a
storage
project
right.
So
we
all
start,
and
now
we
order
by
this
third
column,
and
we
will
get
different
results.
Why?
Because
a
lot
of
people
give
duplicate
stars
by
mistake
or
because
they
want
to
get
a
lot
of
I
know
that
people
are
country.
This
way
it
doesn't
matter
now.
A
A
So
here
I
have
my
query:
just
ran.
Let
me
go
back
to
the
presentation
and
then
not
all
stars
are
equal.
Each
star
has
been
given
by
a
different
person,
a
person
that
maybe
just
knew
to
give
her
a
person
that
might
know
a
lot
of
programming,
a
person
that
knows
a
lot
of
Java
or
JavaScript.
So
each
star
that
I
would
project
get
my
means
something
different.
For
example,
these
are
the
number
of
stars
that
tensorflow
God
last
year,
24,000
versus
prickled
cup.
A
The
row
total.
So
with
this
query,
I
am
looking
at
everyone
that
start
each
of
these
projects,
but
I
also
want
to
know
how
all
they
are.
I,
don't
know
the
age
of
people,
but
I
know
how
old
they
are
there
a
counselor
on
github
or
we
can
see
the,
for
example,
free
code
camp,
the
people
that
start
recode
camp.
They
are
half
the
age,
but
people
that
start
tensorflow
and
they
watch
a
star,
a
1/10
1/8
of
the
projects.
They
left
less
commonly
less
common.
A
Some
repositories
they
have
less
food
requests,
so
we
can
see
that
each
project
has
a
different
structure
of
who
is
starring
it
and
we
might
be
interested
in
different
kind
of
people.
Are
you
interested
in
people
that
are
new
to
github,
or
are
you
interested
of
people
that
have
contributed
a
lot
if
we're
Ana
we're
ranking
again,
just
looking
at
people
starts
from
people
that
have
written
more
than
20
comments,
I
would
rank
exchange.
A
free
code
camp
is
not
long
any
longer
the
top
row.
Yet
our
top
project
is
yarn.
A
People
stirring
with
experience
instead
of
stirring
Facebook
react.
They
start
create
react
up
and
tensorflow,
which
was
a
top-ten
project
on
our
first
ranking.
Now
it's
not
on
the
top
10
I
think
it's
number
11.
The
lesson
here
is
that
we
can
be
the
best
at
anything
if
we
just
find
the
right
ranking
it's.
What
rankings
are
we
interested
in
where
what
what
who's
our
audience
do
we
want
to
speak
to
and
when
you
have
people
starting
your
projects,
you
might
wonder
what
else
do
they
start?
What
else
is
there
interesting?
A
So,
for
example,
with
a
query
like
this
I
will
keep
running
it
for
time,
but
we
list
query:
I'm
taking
everyone
that
start
Esther
flow
I'm,
removing
people
that
start
freaking
come
because
they
start
because
I
decided
to
I'm
free
to
do
it
and
now
I
can
find
that
everyone
that
start
tensorflow.
They
also
start
the
top
early
projects.
They
start
are
also
machine.
Learning
related
and
I
may
be
interested
in
looking
at
those,
and
even
the
result
in
this
case
make
a
lot
of
things.
A
A
This
is
me,
taking
a
picture
of
you
and
I
never
tried
this
before,
but
before
I
started,
I
asked
you
to
fill
a
form
and
in
this
form,
I
was
asking
you
to
tell
me
your
github
logging,
and
these
are
the
responses.
I
got.
Thank
you,
everyone
for
your
responses
and
what
I
can
do
in
bigquery.
I
can
take.
A
A
A
The
top
projects
for
the
people
in
this
room
are
for
2016.
As
soon
as
this
query
runs,
every
programmer
should
now
deep
desire
the
Stars
from
2017.
This
is
what
you
have
start.
Every
programming
should
know
deep
photo
style,
transfer,
grumpy,
I,
don't
know
what
Grampy's,
but
thank
you.
Grumpy
people,
yeah
I
I,
can
look
at
the
history
of
what
you've
done
and
I
can
customize
my
content.
My
talk
I
can
try
to
do
the
best
according
to
what
your
preferences
are,
who
you
are
and
I
can
get
to
know
my
audience
better.
A
Then
we
can
run
from
play.
Timeline
starts
when
do
project
go
out
more
stars.
I
did
with
these
with
their
studios
and
you
free
price
from
Google.
So
here
I'm
looking
at
all
of
the
stars,
the
the
Apache
project-
god,
this
anyone
can
go
to
data
studio,
create
a
dashboard
connected
to
bigquery
and
start
making
it
more
interactive.
For
example,
this
is
a
timeline
of
star
for
Apache
projects
2015,
but
I'm
I
want
to
see
it
by
the
number
of
how
many
stars
each
project.
A
What
so
I
can
do
a
breakdown
by
name
now,
I
can
find
that
this
is
the
timeline
of
stars.
We
can
see
this
huge
peak
here.
Incubator,
but
I
can
even
add
interactive
elements
to
it.
So,
let's
look
at
the
name.
It's
not
expandable
view
and
I
can
publish
this
report
and
anyone
can
come
here
and
I
can
share
it
with
anyone
and
ask
question
like,
for
example,
Apache
arrow,
which
is
involved
these
days.
Let's
only
look
at
Apache
ro
and
we
can
see
the
you've
got
a
lot
of
stars,
February
2016.
A
Then
it
went
down
until
two
weeks
ago
when
they
published
a
new
blog
post.
That
was
super
popular
everywhere,
and
this
is
also
interesting
about
the
Stars
we
most
products
receive
stars
on
pics
not
every
day,
but
yes,
when
they
become
popular,
and
you
might
wonder
how
do
people
find
you
and
the
other
beautiful
queries
that
I
have
more
data
fits
their
shared?
A
You
can
join
those
data
set
and
in
this
case
here
I
charted
three
projects
when
they
receive
stars
and
the
annotations
you
find
there
each
time
this
project
was
featured
on
Hacker
News
on
the
hacker
news
front
page
and
instead
of
running
this
chart
just
by
hand.
Finally,
these
projects
I
was
able
to
find
when
each
probably
lost
our
front
page
just
by
running
this
query,
because
I'm
able
to
join
datasets-
and
let's
look
a
little
bit
on
to
project
health,
what
are
the
healthiest
projects?
What
are
the
projects
that
have
the
most
people
commenting?
A
We
could
find
some
of
these
statistics.
I,
don't
know
if
you
saw
the
state
of
the
October's
today
github
users
bigquery,
to
to
run
many
of
these
queries,
and
let
me
show
you
go
a
little
deeper.
For
example,
this
is
were
in
the
github
is
showing
the
most
discussed
repositories.
What
are
the
repositories
that
got
the
most
comments?
I
can
do
something
similar,
but
I
can
go
deeper.
A
A
These
were
the
projects
that
got
the
most
comments
on
top
kubernetes
17,000
comments,
followed
by
spark
followed
by
openshift,
followed
by
a
mysterious
project
where
I
can
reveal
them
why
they
had
so
many
comments
later
and
the
reason
and
the
other
price
that
you
know
and-
and
you
might
be
still
wondering
why
this
Sauron
day
I'm
up
early
but
I,
never
heard
about
God.
So
many
comments-
maybe
the
most
interesting
stat-
is
not
which
Prague
or
the
most
comments,
but
which
project
has
the
most
people
commented
again.
A
A
A
But
what's
interesting
here
is
not
looking
only
at
how
many
people
are
commenting,
but
at
how
many
comments,
each
people
left
in
the
forum
so
kubernetes
they
have
500
authors
and
each
author
left,
at
least
in
average
18
comments,
the
that's
a
healthy
community.
That's
a
community
where
people
are
coming
back,
documented
they
having
a
dialogue.
Well,
for
example,
in
front
awesome.
They
had
a
lot
of
people
commenting
more
than
8,600,
but
in
average
each
one
left
less
and
two
comments.
This
is
Molly.
Mostly
you
miss
it.
A
A
A
Is
it
possible
I'm
trying
to
people
are
really
nice
on
github.
That's
a
good
thing
in
average.
Yes,
but
what's
even
more
interesting
here
is
not
only
how
people
start
an
issue
which
is
nice,
but
then
how
do
you
get
a
better
closure
rate?
What
issues
get
a
better
closer?
So,
for
example,
it
would
be
nice,
56%,
closure
rate
Versalles.
Is
it
possible
to
have
73%?
A
You
are
being
more
effective
if
you
make
your
request
in
a
more
effective
way.
The
top
one
here
is
I
get
the
following
with
86%
of
closure,
because
you're,
basically
people
being
they're
very
concrete
on
what
they
get,
what
they
want
and
how
they
want
to
proceed,
and
there
is
a
mysterious
so
that
the
humans,
one
maybe
I,
love
math,
is
right.
A
Maybe
the
robots
are
coming
for
us,
but
we
cannot
begin
to
that
later,
but
it's
a
good
mystery
of
why
so
many
authors
have
written
so
that
the
humans,
okay,
other
data
we
can
find
here,
is
where
our
users
are
coming
from
and
instead
of
looking
at
the
the
github
archive
event,
log
we
can
use
the
H
store
and
to
get
that
data,
so
I
will
jump
in
the
query.
I
will
find
all
of
my
users
which
live
in
the
gh
Turin
project.
You
serve
query
table
this
one.
A
We
can
group
by
the
country
code
or
there
by
one
descending,
and
this
is
how
we
can
find
what
are
the
countries
with
the
most
users,
of
course,
on
top?
If
no,
because
not
everyone
puts
their
country
on
the
profile,
but
then
we
have
the
USA
India
China
a
Great
Britain,
but
we
don't
want
to
stop
here.
This
is
just
static
list
of
users.
We
might
want
to
find
the
active
users,
so
we
just
wanna
join.
A
A
Where
are
all
the
people
who
send
in
code
or
interacting
with
github
at
and
it's
taking
a
second,
let
me
continue
with
the
presentation
and
I'll
come
back
to
the
to
the
query,
so
the
top
countries
this
is
our
github
uses
the
country,
but
then
we
can,
for
example,
we
want
to
see
what
are
the
country
sending
the
most
pushes.
Of
course,
the
US
has
a
very
active
users
and
a
lot
of
users,
but
this
is
not
very
fair.
A
We
might
want
to
see
the
number
of
pushes
per
capita,
let's
divide
by
the
population
of
each
country.
So
if
you
divide
by
population
the
u.s.-
yes,
it's
still
one
of
the
top
countries,
but
we
saw
we
see
a
lot
of
programmers
concentration
of
programmers
in
north
Europe.
The
ranking
that
we
get
here
is
that
Iceland
is
the
top
country
just
be
by.
A
We
can
run
another
joint
in
this
case.
I
also
have
a
storm.
Thank
you
all
of
North,
Dayton
week
we
read
this
is
the
this:
is
the
average
temperature
for
each
station
on
each
country
and
this
kind
of
make
sense.
They
are
called
the
countries
and
others
some
have
a
wider
diversity
of
temperatures,
but
at
the
end
we
can
find
a
point,
an
average
point
for
each
country,
and
now
we
can
charge
these
things
together
to
find
that
if
you
look
at
this
chart
on
that
side,
we
have
the
hot
countries
on
this
side.
A
We
have
the
cold
ones.
On
top,
we
have
the
highest
concentration
of
programmers,
the
lowest
concentration,
and
we
can
find
out
correlation
here.
The
coldest
countries
Iceland
there
on
top
have
the
largest
concentration
of
programmers
and,
on
the
other
side,
Africa
is
very
hot
and
has
a
low
concentration
of
programmers
and
an
interesting
place
is
New
Zealand,
which
has
very
high
concentration
of
coders,
but
it's
not
that
cold.
So
it's
a
good
place
that
you
could
move.
If
you
look
looking
for
a
better
weather.
Yes,
you
agree
excellent,
also
interesting.
A
In
Asia
we
have
the
opposite
correlation
at
least
visually.
The
hottest
countries
have
the
most
programmers,
especially
Singapore,
basically
coders,
don't
like
to
go
outside
when
it's
too
cold
or
too
hot
just
to
see
the
last
part
is:
where
do
we
find
all
of
the
code
that
I
promises
you
if
we
want
to
go
and
analyze
the
code
that
people
are
writing?
How
are
the
trends
changing?
We
have
this
new
table
that
we
opened
here.
A
Yes,
June
29.
We
started
sharing
this
repository
of
git
a
copy
of
github
inside
the
query,
so
you
could
analyze
it
when
I
took
the
expression,
this
table
had
200
million
rows
more
almost
2
terabytes
of
data,
it's
even
bigger.
Now
let
me
find
the
current
table,
which
is
github
repos
I
have
contents
right
now
we
have
two
terabytes
of
data,
240
million
files.
A
Now,
if
you
look
at
the
structure
of
this
the
schema
for
each
file,
we
have
the
size.
We
have
the
content
and
the
number
of
copies,
because
duplicate
files
were
only
storing
once.
If
you
want
to
know
how
many
file
this
really
represents.
You
can
sum
the
number
of
copies.
If
you
want
to
know
what
the
number
of
terabytes
we
really
have,
we
can
multiply
the
contents
by
the
size
in
a
few
seconds.
A
As
soon
as
this
run,
we
can
find
that,
yes,
we
have
like
50,
these
files
represent
50
terabytes
of
code,
and
anyone
can
load
this
table
and
start
analyzing.
It
now
be
careful
with
it,
because
everyone
gets
one
three
terabyte
every
month
to
analyze
with
bigquery
and
these
tables
bigger
than
that.
So
I
also
created
a
lot
of
extract,
so
you
can
start
analyzing
Java
code
or
Java
Script
code
or
donate
whatever
you
are
interested
in.
A
So
if
these
are
the
rules,
when
you
want
to
analyze
this
table,
it
has
all
of
the
text.
Files
will
have
less
than
a
megabyte.
Each
file
only
has
one
unique
file
and
copy
you
can
find
out
all
of
the
path
for
each
of
these
files
try
to
not
run
that
joint
to
get
all
of
the
deduplicated,
because
now
you
will
have
duplicated,
because
now
you
will
have
50
terabytes
of
code.
Extract
first
extract
once
use
one
of
the
strikes
I've
an
asked
me
to
do
a
new
strike
go
to
the
account.
A
There
is
a
small
table
with
a
sample
of
contents
and
an
important
fact
here.
We
only
copy
the
projects
that
appropriately
officers
where
we
can
find
that
they
have
an
open-source
license,
so
make
sure
that
your
projects
have
an
open-source
license.
So
we
can
look
at
them
some
example
things
we
can.
For
example,
we
can
go
and
look
at
all
of
the
Java
files,
all
of
the
Java
files
from
2013
or
the
Java
files
from
2016
and
see
what
imports
have
grown
more.
This
were
my
results.
A
A
We're
gonna
start
linking
other
datasets
I
also
have
all
of
Stack
Overflow
in
bigquery.
So,
for
example,
I
can
look
at
what
are
the
top
questions
that
people
are
linking
from
code
to
stack
overflow.
This
is
a
query
run
basically,
I
go
looking
for
any
link
to
Stack
Overflow,
then
I
can
join
with
the
Stack
Overflow
data
set
and
get
the
top
questions
in
the
JavaScript
world.
Everyone
wants
to
know.
There
are
1400
files
that
were
to
know
if
there
is
a
Reigate
xscape
function
in
JavaScript
and
they
link
to
the
Stack
Overflow
answer
now.
A
A
It's
really
interesting.
Now,
what's
really
impressive,
for
me
is
what
unball
test
did
he
wanted
to
know
how
many
people
are
called
copying
code
from
Stack,
Overflow
I'm?
Sure
no
one
here
has
done
that,
but
maybe
we
can
find.
If
someone
has,
he
told
one
of
the
top
Stack
Overflow
questions.
That's
the
answer:
how
to
convert
by
size
into
human,
readable
format
in
Java,
and
now,
if
you
search
for
the
any
text,
just
if
you
just
try
to
search
for
identical
copies
of
this,
probably
people
will
have
changed.
The
formatting
then
bearable
names.
A
So
he
just
reduced
that
piece
of
code
to
a
long,
regular
expression
and
with
the
query
he
was
able
to
run.
These
analyses
find
anything
that
matches
the
regular
expression
and
he
found
448
this
in
Java
files,
we're
copying
the
snippet
and
only
27%,
we're
giving
appropriate
credit.
So
if
you're
popping
code
from
Stack
Overflow
at
least
leave
some
credit
because
they
might
find
what
you
did
tell
your
friends
requesting
features
when
you
want
to
request
new
features,
you
want
to
be
more
effective.
A
You
want
to
ask
them
in
a
nice
way,
for
example,
at
time
until
2
they
go
language
here,
Sam
was
proposing
that
instead
of
writing
after
expiration
time
just
yet,
he
wanted
to
get
an
after
time
until
nice
request.
He
started
it
in
a
nice
way.
I
would
like
to
propose,
but
it's
not
enough
to
do
as
nicely.
A
You
might
also
want
some
data
in
this
case,
my
friend
pre-made
francisc,
went
and
found
every
place
that
would
benefit
from
this
new
function
if
go
implemented
is,
and
he
found
that
more
than
2000
repositories
would
benefit
from
this
feature.
So,
basically,
if
you
put
your
code
on
github,
if
people
are
able
to
see
your
code
and
analyze
it,
your
code
counts
as
a
vote
of
things
that
could
be
better.
The
go
team
went
and
implemented
this
before
a
project.
This
is
our
feature.
A
A
I
wrote
a
long
post
on
how
to
run
a
static,
JavaScript
code
analyzer
inside
the
query.
A
big
query
can
run
arbitrary
JavaScript
code,
so
I
just
found
a
scheme
that
analyzes
JavaScript
code
I
ran
in
his
epic
query.
Inside
my
period
running
the
actual
JavaScript
code,
loading,
a
huge
library
and
I
can
get
answers
like
what
are
the
top
mistakes
:
variables
that
have
not
defined
these
are
warnings
that
things
that
work
on
JavaScript,
but
still
can
we
can
generate
warnings
by
doing
static
code.
Analysis,
I
wrote
this
one.
A
Last
year
we
had
a
lot
of
youth
path
versus
spaces,
who
is
a
tough
person
at
one
spaces?
Okay,
yes,
so
I
run
the
same.
The
same
survey
with
cold
I
lose
the
top
400,000
github
repositories,
so
not
any
piece
of
code
counts,
I,
just
look
at
the
repositories
with
the
most
stars,
I
thought:
I
looked
at
four
big
files,
one
both
were
unique
files
with
my
rules.
A
A
You
can
find
this
do
this
fun
analysis,
but
you
can
even
go
deeper
like,
for
example,
the
operation
rosscup
team.
In
Google
found
that
there
was
a
big
vulnerability
and
using
bigquery,
they
found
thousands
of
open-source
projects
that
had
a
huge
vulnerability
and
they
were
sending
and
they
started
sending
patches
to
everyone
and
yeah.
We
enabled
the
massive
fixes
of
huge
problems.
A
A
Let's
look
at
code
who,
because
people
that
read
lead
in
comas,
they
say
that
they
are
more
productive,
but
if
you
run
over
all
the
sequel
code
on
github,
you
will
find
that
most
people,
as
in
this
room,
like
use,
trailing
commas.
They
put
the
commas
at
the
end.
Leaving
is
is
way
less
people
use
meeting
commas,
but
I'm,
not
FML,
leaving
comma
person
I
find
they're
more
productive.
I
was
not
I.
Didn't
want
to
stop
with
this
and
I.
Don't
want
to
run
a
popularity
contest.
I
want
to
know
which
projects
are
more
successful.
A
I
want
to
justify
my
chest.
So
how
do
you
know
what
rows
are
more
successful?
What
questions
can
you
ask?
The
price
I've
got
the
most
stars,
the
more
stars.
Last
year,
more
contributors,
more
activity
and
joining
these
two
data
set.
I
wrote
this
huge
query
looking
at
all
of
these
metrics
and
it
turns
out
projects
the
trailing
comma,
no,
no,
no,
they
have
in
average
13
stars
project
we
living
or
with
both
have
double
the
number
for
favorite
stars.
Yes,.
A
In
2017,
in
all-time
in
number
of
contributors
in
activity
total
activity
it
turns
out
there
is
a
signal
you
can
use
to
justify
your
usage
of
leading
Commerce
so
who
wants
to
analyze,
github
I
hope.
All
of
you.
Thank
you
for
raising
your
hand,
I,
don't
know
if
we
have
time
for
questions,
but
if
we
don't
have
it,
we
can.
You
can
join
me
later.
I.
A
Have
this
great
video
with
Alex
Allah,
give
her
data
scientist,
because
she
is
telling
me
the
story
of
how
she
did
the
github
October's
last
year,
it's
a
github
and
there's
a
lot
more
of
interesting
analysis.
People
have
been
doing
they.
If
you
do
anything,
please
send
it
to
me
and
keeping
the
full
collection,
and
if
you
have
any
questions,
find
me
on
reddit
find
me
on
Stack
Overflow
from
me
on
Twitter
and
if
you
have
any
feedback
for
me,
I'll
a
feedback.
So
thank
you
very
much.