►
From YouTube: 2019-10-10 Python SIG
Description
No description was provided for this meeting.
If this is YOUR meeting, an easy way to fix this is to add a description to your video, wherever mtngs.io found it (probably YouTube).
A
A
A
C
D
A
A
C
C
A
E
C
A
C
F
F
Yeah,
not
a
lot
to
say,
I
mean
once
we
fix
these
things,
and
there
is
a
couple
of
two
dudes
that
are
kind
of
optional.
I
mean
I,
wasn't
sure
whether
it
was
worth
delaying
the
PR
further,
for
example,
until
implement
the
inject
and
extract
methods,
but
yeah.
So
at
any
point
we
can
decide.
Okay,
this
is
good
enough.
We
can
merge
it
now
and
then
you
know
break
down
the
missing
things
into
separate
PRS
or
wait
until
everything
is
implemented.
So
I'm
fine
with
both
and.
A
F
A
F
Sure
the
biggest
part,
like
the
biggest
part
in
feedback
that
I've
received
up
to
now,
besides,
you
know,
trickle
stuff,
like
moving
things
to
a
separate
directory.
I
was
from
Christian
thanks
a
lot
it's
regarding
the
handling
of
context
managers.
I
had
made
some
parts
of
that
I
think
these
are
some
parts
of
implementation
that
were
a
little
fuzzy
in
my
head,
so
I
parently
have
done
some
things,
that
kind
of
maybe
work
by
accident,
but
don't
make
a
lot
of
sense.
F
So
this
is
like
the
biggest
part
that
I'm
currently
you're
factoring
yeah
other
than
that
I
think
I'm,
pretty
comfortable
with
implementation.
One
thing
I
haven't
done
yet
is
actually
try
to
run
some
reliable
test
bed
or
some
actual
tracing.
Besides,
you
know
a
tiny
demo
application
to
see
that
this
is
actually
doing
things
that
make
sense.
Maybe
I
can
also
use
the
yogurts
order
to
look
things
in
iaeger
or
in
lights
or
whatever,
just
to
see
that
visually
things.
F
A
By
the
way
for
the
4d,
for
these
grab,
your
library
being
on
the
top
level
or
not
in
Java,
we
have
it
as
top
level,
but
it's
mostly
because
we
plan
to
move
it
outside
that
repo
eventually,
but
we
are
having
it
there.
You
know
because
it's
easier
for
us
to
test
things
right
there,
but
the
plan
is
to
move
it
out
outside
so
prowling
with
the
bottom.
We
do
this.
Despite
own
layer,
we
will
probably
do
the
same
I
think
it
doesn't
belong
there,
although
it
for
now
it's
it's.
Okay,.
C
A
Well,
I
remember
having
this
talk
with
Bogdan,
at
least
for
Java.
We
decided
because
we
were
afraid
that
the
code
base
would
be
to
be.
You
know
first
and
then
it's
too
specific
to
open
Tracy.
So
he
was
saying
that
probably
like
no
people
who
had
a
lot
of
opposites
background
wouldn't
be
so
you
know
affected
by
death
and
it
could
be
noise
but
yeah
we
can
discuss
if
you
guys
think
it's
better
to
keep
it
here.
We
can
also
consider
consider
death,
okay,.
C
I
mean
because
we
we
have
two
options.
We
can
either
like
put
things
in
a
separate
like
package
or
like
we're
doing
now,
or
we
can
put
it
in
a
totally
separate
repo,
and
so
we've
gone
the
route
with
just
like
having
a
bunch
different
artifacts,
but
we're
gonna
move
the
the
third-party
exporters
out
and
separate
repos
anyway,
so
yeah
we
will
have
to
yeah.
F
Yeah
yeah
I
mean
usually
the
things
I'm.
The
most
worried
about
is
dependency
management
when
we
discuss
this
kind
of
stuff,
like
one
repo,
multiple
repository
in
this
case,
I
mean
we're
installing
everything
in
a
big
way,
and
we
give
the
full
fully
qualified
package
name
even
right
now,
when
we
have
everything
in
a
single
repository,
so
I
don't
think
it
matters
too
much
in
terms
of
in
terms
of
dependency,
it's
just
a
matter
of
like
how
to
manage
the
source
itself
and
not
really
how
to
install
or
reference
dependencies
between
the
components.
A
A
So
you
know
how
do
you
guys
know?
The
current
name
is
start
as
Greenspan.
Well,
we
have
two
suggestions.
One
is
from
to
use
start
a
new
span
and
the
other
is
to
discover
a
user
span
direct
in
the
method.
So
please
just
leave
a
comment
to
the
PR
or
feel
free
to
discuss
it.
Now
here,
yes,
I'm
fine,
either
way.
A
C
This
is
because
I
looked
at
the
the
ticket
in
specs.
We
have
kind
of
a
weird
requirement
now,
where,
before
we
had
a
single
global
tracer,
now
we
will
have
something
like
a
single
global
tracer
factory,
and
then
we
will
set
like
the
config
that
used
to
be
set
on
the
tracer
on
the
trace
factory,
and
then
the
idea
is
that
users
that
get
the
factory
instead
of
the
tracer
can
write
at
an
it
time,
use
the
factory
to
get
one
or
more
tracers
and
encash
them.
B
Like
the
the
tracer
will
then
only
have
will
be
a
quite
lightweight
object.
It
will
probably
only
have
instrument
instrumentation
that
will
rename
and
version
these
two
parameters
you
pass
to
the
factory.
The
reference
to
the
factory
and
the
factory
has
tech,
chure
data
like
the
spam
process,
so
stay
active
span
and
so
on.
B
F
B
There
is
the
expect
currently
says
that
it
is
required
that
you
specify
the
name.
This
was
yeah.
This
was
the
result
of
the
discussion
it
is
required,
but
due
to
the
way
errors
are
handled,
there
will
be
effectively.
There
will
be
an
default
tracer,
because
when
you
don't
specify
a
name,
it
is
an
error,
but
we
will
handle
that
error
by
giving
you
some
default
tracer.
That
also
works.
B
C
Yeah
so
I
I'm
interested
in
working
on
this
cuz
I
started
to
look
at
not
logging
and
I
was.
You
know,
curious
how
we're
gonna
treat
a
log
correlation
and
first
class
loggers
eventually
and
open
telemetry,
so
just
like
looking
out
to
the
future,
I
would
like
to
you
know,
get
it
to
get
a
sense
for
how
a
logging
loader
might
work
so
Christian.
If
you
want
to
work
together
on
this,
and
have
it
pick
this
up
with
you,
yeah.
B
G
Don't
have
a
lot
of
context
on
this
tip
yet,
but
I'm
just
curious.
You
know
considering
a
blog
correlation
as
a
goal.
Does
it
make
sense
to
maybe
just
have
a
global
object
that
is
namespace.
That
then
has
its
own,
like
tracer,
meter
and
potentially
Locker
object,
ensure
that
that
correlation
is
kind
of
tightly
coupled.
So
the.
C
Way
that
that
long
correlation
will
work,
at
least
at
first,
is
that
there
is
no
like
there's
no
longer
at
all.
As
far
as
open
telemetry
is
concerned,
it's
just
that
when
you
use
like
the
built-in
Python
loggers
you'll
get
some
attributes
from
the
current
span
right.
So
you've
got
like
the
trace,
ID
and
span
ID
and
trace
options.
G
Yeah
and
I
think
that
makes
sense,
I'm
just
curious
with
it.
Well
I
guess
the
logger
call
would
be
executed
within
the
context
manager
and
this
that
would
that
would
give
you
I
guess
the
the
tracer
that
you'd
be
using
in
question
ago,
though
I
guess
yeah,
maybe
I'll
just
read
through
this
spec
a
little
bit
more.
C
The
other
good
thing
about
this
and
the
reason
I
don't
think
we're
gonna
have
at
least
at
first
a
like
a
single
global
container
for
for
all
of
this
limiter
stuff.
Is
that
we're
gonna
have
traces
scopes
to
requests
and
then
we're
gonna
have
metrics
like
created,
statically,
probably
and
it'll,
be
you
know,
per
class
or
module
or
something
so
you
know
we
may
start
a
different
name
tracer.
C
G
C
Yeah,
okay,
I,
said,
look
at
the
comment
now.
It's
true
I
think
everybody
is
in
favor
of
PI
test,
or
mostly
everybody,
yeah
I,
think
it's
fun.
So
in
this
comment,
I'm
just
curious
because
it
doesn't
look
like
PI
test
is
because
it
look
like
you're
actually
using
any
features
of
PI
test
in
this
PR
hi.
C
F
C
G
C
D
C
B
C
A
C
E
E
B
C
C
F
E
Yeah
then
gonna
test,
so
it
so
it
can
bring
some
benefits
in
the
future
if
we
allow
or
use
by
tests
or
features,
but
the
the
brownest
for
this
particular
particular
match
request.
Oh
I'm,
sorry
pull
request.
We
are
not
on
github
sorry,
so
for
this
it's
not
so
necessary.
Probably
but
I
used
one
feature
of
Pi
test
that
allows
me
to.
E
To
import
every
Python
file
in
order
to
get
it
listed
in
the
coverage
reports,
because
when
I
was
just
running
the
pie
test,
some
of
the
existing
files
are
not
imported
at
all.
So
I'm,
not
sure
why?
Probably
they
are
not
covered
by
tests
and
they
are
not
used
by
any
models
that
were
covered
by
tests.
So,
as
you
can
see
in
the
PR
for
for
the
coverage
test
and
it
uses.
E
A
F
Summarize
what
you
just
said,
I,
don't
personally
don't
have
a
strong
opinion
like
for
or
against
spiders
I
was
just
asking
the
question,
because
we're
talking
about
introducing
a
new
dependency
of
the
project
and
usually
I
try
to
do
that
when
I
actually
have
a
problem
that
I
need
to
solve
with
that
dependency.
Otherwise,
why
I
add
a
new
component
right?
So
this
is
why
I
was
asking
the
question
just
trying
to
understand
if
we
currently
have
an
issue
that
we're
trying
to
solve
and
pipe
test
as
a
possible
solution
to
it.
Yes,.
F
F
E
C
B
A
lot
of
sense:
okay,
yeah
I,
have
no
strong
opinion
on
this
yeah
I
will
sorry
I
agree
with
Chris
yeah
that
then
that
also
means
that
probably
I
will
create
an
additional
package
that
has
the
shared
test
code,
because
you
wouldn't
want
to
publish
the
test
code
with
normal
whiskey
package.
So
we
have
to
as
your
test
code
package,
but
that's
probably
a
good
idea.
Anyway,
then
we
can
share
a
test
code
between
all
tests,
API,
SDK
and
so
on.
G
You
really
can
you
like
your
prefix
I.
Can
oh
yeah
it's
okay
for
other
people
drop
off,
but
hey
on
the
the
3
C
3
trace
date,
stuff
I've
been
working
I've
been
trying
to
get
that
working.
G
G
G
E
D
A
B
D
B
B
F
B
B
F
G
G
A
D
G
G
Because
there's
a
distinction
there
and
then
yeah,
she
works
well
when
you're,
like
with
an
egg
link
when
you
modify
the
files,
you'll
actually
see
that
reflected
in
whatever
Python
environment
installed
that
as
an
editable.
This.
A
G
This
scenario,
if
you,
if
you
go
and
modify
those
source
files,
they
don't
affect
the
virtual
environment
because
it
was
installed,
he's
right,
copied
over
right,
got
it
great
questions
and
also
it
looks
like
we
have
both
pip
install
ease
and
pip
installed.
Just
relative
directories
for
different
talks
environments,
so
that
might
be
worth
normalizing,
but
yeah.
B
B
G
Yeah
I
mean
I,
don't
wanna
like
bike
on
this
stuff,
but
I
did
terrible
on
there.
Well
yeah,
actually
maybe
but
Dillo.
Basically
I
built
a
tool
that
creates
a
virtual
M&N,
allows
you
to
declare
tasks
using
Python.
You
can
just
execute
those,
and
so
you
know
it.
Programmatically
have
all
of
our
tests
and
Olympian
stuff
executed.
G
E
B
A
G
See
I
mean
actually
I
kind
of
fine
yeah
I
mean
there's
pros
and
cons
to
that
approach.
I,
don't
actually
I
use
PI
n
locally,
so
I,
don't
you
have
I
only
have
one
version
of
Python
install
any
given
times
when
I
run
talks,
there's
like
50
environments
that
actually
don't
execute
yeah
and,
like
you
know,
they
call
that
Travis
does
provide
like
the
ability
to
to
multiplex
your
job
across
multiple
Python
versions.
Oh
nice.
G
Mean
that's
that's
what
I
think
talks
is
expecting
you
to
do
yeah,
I,
very
much
yeah,
and
actually
it
looks
like
we
do
both
here
where
the
Travis
Tamil
is
setting
the
Python
version.
So
it's
sort
of
multiplexing
jobs
against
those,
and
then
we
run
talks
which
selectively
chooses
based
on
the
pythons
that
are
installed,
and
so
we
actually
have
a
bunch
of
I'm
actually
guessing.
If
you
look
at
the
Travis
CI
file,
we'll
see
a
bunch
of
missing
skipped
executions
per
for
environment,
but.