►
From YouTube: Weekly Sync 2020-05-01
Description
Meeting Minutes: https://docs.google.com/document/d/16u9Tev3O0CcUDe2nfikHmrO3Xnd4ASJ45myFgQLpvzM/edit#heading=h.ltd9zn256z85
A
A
A
A
A
A
Sub
spec,
okay,
so
we
need
to
expand
this
to
basically
take
sort
of
arbitrary
data
types
well,
somewhat
arbitrary.
We
can
do
basically
name
tupple
and
things
like
that,
we'll
addictive
dict
and
list
containing
name
tuples
or
data
classes.
We
could
do
that
kind
of
expansion
there,
because
we
have
the
sub
spec
stuff,
which
is
actually
the
other.
No,
that
was
not
the
other
pull
request,
but
there's
another
poll
request
that
went
up
for
that,
but
basically,
what
we're
doing
is
just
making
it
easier
to
define
in
operations.
A
So
if
you
have
any
function
and
you
just
throw
up
on
it
now,
you're
gonna
have
an
operation
and
then
obviously,
what's
going
that
that
that's
going
to
look
like,
is
it's
sort
of
what
we've
already
been
doing?
If
you
don't
have
the
definition
in
the
global
space,
you
can
just
say
you
know,
operation,
you
know
the
the
function
op
and
then
inputs
and
then
whatever
the
input
is
here
and
it'll
show
up
as
that.
A
A
There
should
be
an
issue
up
that,
let's
see
mm,
oh
yes,
we
also
have
an
issue
up
for
basically,
if
there's
only
if
you
put
a
type
annotation
on
and
it's
got
one
single
value
as
a
result,
then
this
is
one
that
somebody
could
jump
on
and
do
then
we'd
create
the
definition
for
the
result,
and
that
way
you
know
if
this
thing
returned
anything
you
could
just
like.
If
we
return
to
bool-
and
you
said
you
know-
arrow
bool
with
the
Python
type
hinting
then
we'd
get.
A
You
know
in
OP
that
outputs
you
result
and
the
definition
would
be
of
primitive
pool,
and
the
other
thing
is
that
when
we
do
this,
this
is
sort
of
leading
into
something
that
we're
going
to
need
to
start
doing.
We've
talked
about
this
a
bit
before
which
is
sort
of
like
this
concept
of
name
spacing.
So
if
you
look
at
like
you
know,
github
we
have
Intel
/,
DF
FML,
so
Intel
is
kind
of
like
the
namespace.
If
you've
done
C++,
you
know
you'd
import
from
STP.
A
Okay,
so
I
create
this
data
as
another
example:
I,
add
it
the
other
day,
but
so
you
create
this
data
flow
and
well.
What
are
all
the
definitions?
Well,
there's
data
to
print
key
mapping
and
value
right.
So,
if
I
have
any
other
operations
that
have
key
mapping
or
value
as
their
definition
name
now
sudden,
it's
overridden.
So
what
this?
What
this
patch
also
did
our
subsequent
patch
out
for
this
one?
A
It
made
it
so
that
when
these
definitions
are
created
on
the
fly,
it
actually
creates
them
in
the
format
of
module,
name,
dot,
function
that
we're
wrapping
dot
input
name
so,
for
example,
DF
FML
time
mapping,
dot,
create
key
ends
up
becoming
DF
FML
mapping
create
key,
and
so
that's
how
that
would
work.
And
that
way
you
know.
If
you
look
in
the
definitions
and
we
had
two
things
where
they
had
key,
then
they
wouldn't
collide.
They
wouldn't
overwrite
each
other.
So
that's
a
thing
that's
happening
and
we're
gonna
need
to
start
doing
that.
A
To
sort
of
you
know
everything
so
that
we
don't
get
collisions.
So
if
you're
creating
new
definitions
to
sort
of
keep
in
mind
that
we
should
probably
return
to
namespace
them,
and
that
means
so
arguments,
I
read
it
on
the
floor
and
put
dick
created,
if
not,
access,
definitions
created
for
each
argument
and
then
keep
in
mind.
A
Okay
and
then,
let's
see,
what
are
we
cover
on
Tuesday
I?
Think
I
did
something
since
Tuesday
that
everybody
needs
to
know
about.
Oh
no,
okay,
we
got
that.
So
if
we
didn't,
if
you
didn't
hear
it
on
Tuesday,
basically
like
for
all
the
examples
and
all
the
docstrings,
we
can
just
do
from
DF
FML
import,
whatever
you
don't
have
to
do
the
full
path
down
with
the
FML
dot,
util,
dot,
async,
test,
cakes
or
whatever.
You
know
that
kind
of
stuff.
A
A
Why
is
this?
Why
is
this
interesting?
It's
because
all
classes
take
a
config
object,
and
so
previously,
if
you
didn't
pass
a
config
object,
even
if
there
was
no
fields
to
set
in
that
config
object,
you'd
have
to
pass
it
some.
You
know
anything
Stan.
She
ate
it
config.
Now
you
can
just
call
or
instantiate
the
object,
and
it
will.
It
knows
that
it
goes
and
checks
that
everything
is
a
default,
and
then
it
just
instantiates
it.
A
B
A
C
A
Okay,
let's
see
yeah,
that's
a
good,
that's
a
good
question.
Well,
I
would
just
sort
of
do
a
check
on
the
file
size
right
now
and
make
sure
it's
like
above
some
value,
because
I
suspect
we
don't
want
to
do
a
hash
of
the.
Let's
see
we've
got
ffmpeg.
Okay,
were
you
able
to
throw?
Does
it
does
it?
I
haven't,
looked
at
the
see
I
recently
did
you
were
you
able
to
put
the
I,
don't
see
anything
that
might
be
doing
that?
Let's
see.
C
A
C
B
A
C
A
Great,
so
this
is
important
that
we
know
this,
though,
because
I
was
messing
with
something
in
the
SQL.
The
custom
SQL
example
and
I
was
like
I.
Don't
I
don't
feel
like
this
test
is
running
and
I
think
we've
concluded
that
that
test
is
not
running.
So,
let's
make
an
issue
for
that,
so
this
issue
was
supposed
to
before
what
was
this
issue
before
you
guys
remember
what
this
is
all
is
this
I
may
not
have
even
said.
A
Okay,
cool
thanks
for
jumping
back
on
or
getting
back
to
me
here
and
this
because
I
was
I,
haven't
been
super.
Looking
at
this
one,
cuz
I
know
it
was
sort
of
a
long-running
thing,
so
it's
just
waiting
for
you
to.
Let
me
know
when
you,
when
you
had
things
that
needed
review
so
perfect.
Thank
you.
Alright
great!
So,
let's
see
who's
up
next.
A
C
A
A
This
thinks
this
build
time
is
so
long
that
I
don't
know
if
there's
a
good
if
anyone's
made
like
a
good
editor
integration
like
a
markdown
or
artist
preview
type
thing,
but
to
do
the
full
Sphinx,
so
yeah
I
don't
know
it
might.
This
is
not
it's
it's
not
great.
That's
not
the
best
workflow
there,
so
yeah
I!
Guess
if
you
find
anything
out.
If
anyone
finds
anything
out,
let
the
rest
of
us
know.
A
Okay,
let
me
see
I
may
have
got
two
computers
open
here,
so
let
me
just
go
to
the
other
one.
Real,
quick,
I
think
I
may
not
have
hit
the
review
finished
button
on
that
one.
Let's
see,
that's
just
looking
at
that
open,
Oh
goddamn
if
I
did
not.
This
always
happens
to
me.
Sevens
me
a
lot
if
you
guys
are
like
waiting
for
me
to
review
your
stuff
quite
likely.
I
have
forgotten
to
hit
the
submit
review
button.
A
So
sorry
about
that,
just
always
just
ping
me
and
be
like
hey
what
happened?
It's
usually
forgot
to
hit
a
button,
okay
and
it
that
okay,
yes,
this
I
mean
you're
definitely
on
the
right
track
here.
So
that's
good
good
job
with
this
you're
on
the
right
track,
I
mean
the
review
sort
of
says
all
I
have
to
say
about
this,
but
basically
you
don't
really
need
these
pi
PI
operations.
A
D
B
A
A
A
A
This
is
just
you
know,
it's
just
sort
of
rinse
and
repeat
right:
we've
get,
we've
got
we'll,
have
JavaScript
in
here
and
we've
got
Python
and
then
we'll
probably
want
to
tweak
python
a
little
bit
because,
basically
right
now
it
clones
the
guy
repo,
it
grabs
the
name
and
then
it
reruns
old
stuff,
which
then
goes
and
downloads
the
latest
package
and
runs
bandit
on
it
right.
So
then
you're
not
actually
running
branded
on
the
right
thing,
so
we
probably
just
want
to
run
bandit
right
there.
B
A
B
A
B
E
A
E
A
A
A
A
A
A
A
A
Okay,
I'm
just
thinking,
okay,
so
I'm
thinking
what
I'm
thinking
here
is.
We
can
have
one
operation
to
it.
Just
like
multiply
something
by
ten
right
and
then
we
just
take
all
the
features
and
we
say
run
them
through
the
multiply
by
ten
operation.
And
then
the
question
is
well:
how
do
we
get
them
back
out
right
and
into
their
their
previous
locations?.
A
A
A
A
A
A
A
So
in
that
case
we
could
do
feature
name
to
actually
that's
perfect.
Yeah.
That's
is
perfect.
Okay,
so
we
can
use
this
associate
spec.
So
this
one
Maps
file
name,
you
know
boolean
value
right.
So
we
end
up
with.
Is
this
file
path
and
then
the
pool
and
what
we
want
is
the
feature
name
and
then
the
result
of
the
feature
times
ten
right.
A
A
A
C
A
Okay,
you
know
I'm
gonna
get
back
to
you
on
this
one.
There's
a
way
to
do
this.
I
swear.
There's
a
way
to
do
this.
C
A
A
A
A
A
Do
we
have
a
way
to
say
it
can
come
from
multiple
definitions?
Don't
think
we
have
a
way
to
say
can
will
come
from
multiple
definitions
and
that's
what
we
need
right
now
is.
We
need
a
way
for
an
input
to
say
that,
because
so
for
the
flow
right
now
what
the
flow
structure
does
is.
It
says
we
define
where
the
inputs
are
allowed
to
count
like
where
each
input,
let
me
pull
up
an
example.
A
So
the
flow
currently,
basically
it
says:
where
is
the
input
allowed
to
come
from
so
this
for
print
output?
The
data
input
is
allowed
to
come
from
the
mapping
output
of
DF
of
Multan
mapping
to
create
and
for
DF
Melda
mapping
to
create
the
key
input
is
allowed
to
come
from
any
of
this
any
inputs
in
seed
there
with
a
matching
definition
and
the
value
is
allowed
to
come
from
any
inputs
and
seed
which
are
math,
which,
with
a
matching
definition
of
matching
definition,
is
value
or
key.
A
A
A
A
A
So
we'd
have
just
so
I'll
just
agree
whether
this
makes
sense
or
not
so
from
the
origin
seed.
We
can
accept
definitions
in
the
format
of
let's
see
right
from
the
we're.
Basically,
what
we're
seeing
here
is
for
the
and
let's
make
it
applicable
to
edit
feature
here
so
for
the
flow
for
edit
feature.
The
inputs,
which
is
where'd
you
go
feature
feature,
can
come
from
seed
and
within
seed.
It's
allowed
to
be
years.
A
E
A
So
what
I'm
saying
is
that
the
within
a
day
of
the
dataflow
structure,
we
have
this
flow
property
and
the
flow
property.
Is
this
dict,
where
the
keys
in
the
deck
are
the
operation
names
or
the
operation
instance?
Names
is
what
they
really
are,
and
so
this
edit
feature
operation
it
that
within
flow
it
has
this
sub.
It
has
this,
it
so
flow
is
a
dict
and
then
the
keys
in
the
dict
are
the
operations
that
are
in
your
data
flow.
Now
each
operation
is
also
a
dict
and
it
has
inputs
in
it.
A
D
A
So,
if
I'm
about
to
run
edit,
if
I'm
the
the
thing
that
runs
the
data
flow
and
I
need
to
decide.
Okay,
like
I,
have
these
inputs
entering
the
entering
the
network
right,
so
I'm,
given
inputs
and
I
need
to
decide
what
operations
do
these
inputs
go
to
I,
consult
the
flow
and
I
say:
okay,
this
this,
for
example,
we're
going
to
add
some
inputs
and
we're
going
to
add
an
input
with
a
definition
of
years,
trust,
experience
and
salary
right.
Those
are
the
that's
that's.
A
What
we've
done
here
is
we've
said:
okay,
this
source
is
going
to
add
inputs
which,
which
definition
names
match
the
feature
names
right?
Yes,
okay,
so
we
want
these
definitions
which,
by
default,
all
inputs
are
from
the
seed
origin,
and
this
is
something
that
I
haven't
really
explained
really
well,
but
the
idea
behind
this
is
that
you
could
have
sort
of
different
origins
of
inputs.
A
We've
ensured
that
nothing
from
like.
So
if
this
said,
for
example,
trusted,
then
only
things
that
came
in
through
a
trusted
source
and
were
tagged
trusted
would
be
allowed
to
go
into
in
would
be
allowed
as
an
argument
to
edit
feature
right.
So
if
my
client
sent
me
something
that
always
a
matching
definition,
it
wouldn't
be,
it
would
never
go
to
edit
feature
right
so
because
we
don't
trust
it
that
the
source
is
not
trusted
so
or
the
origin
is
not
trusted.
E
A
Then
the
the
other
thing
that
you'll
see
some
like
the
other
way
of
specifying
this
is
that
we
have
basically
this
single
pair
here,
which
would
be
saying
you
know
if
we
had
this
other.
Let
me
add
this
here,
and
this
is
it's
obviously.
This
is.
This
is
also
why
we
have
that
oops.
This
is.
This
is
also
why
we
have
that
diagram
command
is
so
that
we
can
visualize
this
whole
thing.
A
So,
for
example,
now
this
syntax
says:
if
there's
not
a
list
here
basically,
so
if
this
is
not
a
list,
then
this
is
the
specific
input
from
this
operation.
If
this
is
a
list,
then
we're
going
to
treat
it
as
these
are
definitions
from
this
origin,
and
that's
something
we're
going
to
have
to
go.
Do
so
I'm
gonna
make
a
screenshot
of
this.
A
A
A
D
Okay,
okay,.
A
A
A
A
A
That's
probably
that
just
the
case
that
we
need
to
write
another
output
operation,
it
can
be
very
much
like
associate
it's.
The
beauty
of
operations.
I
was
asked.
I
was
trying
to
come
up
with
this
stuff.
Like
a
couple
years
ago
now
and
I
was
at
my
work
and
I
asked
one
that
I
was
like
it's
really
stuck
I'm
like
how
do
we
get
the
data
out
of
the
network
like
what?
How
do
we
do
that
right,
like
we
wind
up
with
all
this
data
in
the
network,
and
what
do
we
do?
A
We've
got
all
these
operations
and
I
was
explaining
it
to
this
very
senior
guy
and
he
was
like
well
just
make
got
operations
just
make
more
operations.
I,
don't
think!
That's
brilliant
I'll
just
make
more
operations,
no
give
it
output
operations.
Okay,
let's
see
by
origin
by
origin
by
origin,
where's
flow
input,
flow
inputs,
update.
A
So
if
anyone
didn't
catch
me
like
if
anyone
needed
something
I'm
gonna
have
to
ask
you
to
please
ping
me
on
get
her,
because
I
have
to
make
sure
my
review
stuff
is
all
very
solid,
because
this
is
the
final
day.
So,
let's
see
import
input
flow
and
put
flow
and
put
flow
from
dick.
So
that's
going
to
be
critical.
Okay,.
A
A
A
A
A
A
A
E
A
A
Yeah,
you
probably
want
to
create
a
new
output
up.
You
can
work
on
the
output
operation
right
now,
output
operation,
which
Matt,
which
is
similar
to
associate
let's
see
you're
basically
gonna,
take
feature
name
equals.
Let's
see
what
is
that
associated
feature
name
and
then
yeah?
It's
similar
to
associate.
A
A
E
A
A
E
A
Yeah
and
you
might
might
yeah
take
pixie-
do
this
output
operation
first
so
that
you
can,
because
what
you
can
do
here
is
you
can
just
throw
the
inputs
in
the
network
and
then
have
you
know
to
minimally
to
test
it.
A
You
can
create
one
operation
each
right
and
then
ensure
that,
at
all
that
your
output
operation
is
working
there
and
then
circle
back
to
this
guy
I
mean
I'd
love
for
you
to
understand
this
more
because
that's
always
helpful
for
more
people
to
understand
the
dataflow
stuff,
but
it
might,
it
might
be
kind
of
funky,
but
so
it
might
take
a
while
but
yeah.
If
you
can,
if
you
want
to
dive
into
that,
that
would
be
awesome.
So
all
right,
I'm,
sorry
guys,
I'm!
A
A
Pr
on
okay,
so
I've
got
I've,
got
notes
on
who
went
today
and
then
whoever
didn't
go
today,
we'll
we'll
talk
on
on
Tuesday
then,
and
if
you
want
to
talk
to
me
in
between
then
just
just.
Let
me
know
I'll
get
her
or
we
could
have
a
meeting
so
All
Right
see
you
guys.
Did
anybody
have
anything
sort
of
urgent
or
is
that
work
for
everyone.