►
From YouTube: SunPy Developer Discussion
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
Don't
know,
there's
not
a
lot,
we
can
do.
To
be
honest,
I
mean
the
thing
is
we
can
switch
from
hashes
to
whole
files
if
we
wanted,
but
if
one
of
the
dependencies
change
still
going
to
cause,
although
Tom
are
has
a
package,
he
calls
hi
test
mvl,
which
stores
the
full
figures,
and
then
it
can.
It
has
a
tolerance
on
the
image.
So
if
it
only
changes
a
little
bit,
the
test
doesn't
favor.
That
involves
storing
the
whole
image
in
the
repo,
which
is
why
we
came
up
the
vicar
hashes
idea.
C
A
A
A
A
Fair
enough
right,
yeah,
there's
that
pull
request
I
made
that
path,
lakes
that
make
a
hash
is,
but
the
windows
built
and
broken.
So,
let's
see
whether
I
just
restarted
the
windows
build
we'll
see
where
that
comes
back.
Hopefully
that
will
fix
it,
and
then
we
can
fix
everything
else
with
it.
A
A
So
if
you
delete
the
database
file
and
then
rerun
the
tests,
it
will
recreate
you
a
new
database
file
that
you
then
copy
into
the
right
place,
and
the
challenge
is
just
making
sure
that
they're
right
when
you
do
that,
because
you
don't
get
to
see
the
plot,
which
is
something
I'd
like
to
change
but
I,
think
that
if
they
plot,
if
you
don't
have
a
if
there's
no
hash
database,
it
should
display
you
the
pot
so
that
you
can
check
the
right
and
I.
Don't
know
how
to
do
that.
A
A
A
A
C
A
A
That's
the
thing
at
the
moment
we
have
we,
we
saved
the
plots
in
the
tests,
say
about
a
PNG
file,
and
then
we
compare
the
hatch
of
that
PNG
file
to
a
reference
half,
and
we
do
it
like
that,
so
that
we
don't
store
all
the
pictures
in
the
repository
and
blow
the
repo
up
with
all
the
pictures
as
they
change
and
but
that
has
some
downsides
like
any
tiny
change
to
any
pixel
at
the
PNG
kind
of
breaks
test.
There's
no
way
to
building
a
tolerance
when
you
are
restoring
a
hatch
and
is.
A
Yeah,
maybe
that
and
but
the
moment,
yeah
that's
what
I
was
saying
basically
is
there
needs
to
be
a
way
to
compare
the
plots,
or
at
least
not
even
maybe
even
if
not
compare
the
box
just
see
the
new
one.
So
you
can
go.
A
lot
looks
like
a
plot.
It
hasn't
gone
completely
mad,
although
preparing
the
plots
would
be
nice,
but
I
would
involve
running
a
tests
on
mobile
for
committees
from
PI
test,
and
that
would
be
interesting.
C
A
A
C
A
A
A
A
C
A
A
D
A
B
A
A
B
C
B
With
the
original
title
of
the
SCP,
as
I
understand
it,
was
to
make
light
curve
very
similar
to
map
whatever
to
work.
Very
simply,
the
EP
is
very
similar
for
users
and
also
that
to
them
stuff,
like
the
downloading
files,
can
be
used
in
a
very
similar
way
for
both
the
benefit.
To
that
word,
easy
easy
to
understand
and
maintained
across
the
two
different
classes.
Right
am
I,
describing
it
well
enough,
then
yeah.
B
You
know
what
I've
got
now
free
here
in
there,
and
so
that's
why
I
think
I
kind
of
asked
about
it,
but
this
as
I
understand
map
you
have
to
put
in
a
list
of
files
or
filename
totally
actually
and
my
objection
that
that,
as
far
as
a
light,
curves
concern
is
that
the
light
curve
user,
both
from
what
I'm
already
used
to
in
Python
a
lot
of
people,
are
used
to
in
audio
letter
W.
Is
that
you
don't
necessarily
just
pick
a
file
name.
A
B
B
Way
we
can
talk
about
all
this
is
imminent
fizzle
to
happen
soon
and,
of
course,
it
isn't
drags
on
so
like
if
I
see
something.
Oh,
that
would
be
good
for
likable.
That
would
be
good
for
they
specifically
a
go
like
a
more
specifically
a
liar
like
us,
a
nice
functionality
or
some
or
even
some
bug.
You
know
me
as
an
example
would
be
far
less
likely
to
put
a
huge
amount.
A
A
C
D
C
Still
think
that's
a
good
idea:
I,
don't
like
the
idea
of
taken
in
of
a
time
range
and
a
measure.
Sorry
a
spacecraft
or
an
instrument,
or
anything
like
that.
I
think
following
the
map
methodology
is
the
way
to
go.
Who
makes
things
simpler
to
maintain
and
one
side,
no,
the
lake
of
itself
can
can
be
abstracted
out
from
the
damn
building
retained
again
in
the
future.
So
I
I
am
a
fan
of
the
of
the
map
method.
C
B
C
Why
is
that
cumbersome?
They
make
one
request
for
like
a
week's
worth
of
data
of
months
worth
of
data
and
I
got
a
file
list
and
then
apply
past
that
file
list
into
the
lake
of
constructor,
and
then
that
the
like
of
constructor
handles
the
actual
construction
of
the
lake
of
taking
into
ken
gaps
potential
gaps.
Then
I
think
that's
a.
C
C
So
when
I'm
constructing
a
like
a
map
cute
for
example,
then
I
don't
like
I
commonly
want
to
know
what
the
what
every
file
is
right,
but
I
have
a
list
of
files
enemy
that
I
can
examine
later
on.
There's
nothing
mysterious
like
if
some
of
them
double
up
all
of
those
one
or
two
missing,
then
I
of
that
list
of
files
enemy.
If
you
try
and
combine
the
acquisition
of
data
with
its
presentation
and
storage
and
another
an
object,
then
kind
of
confused
those
two,
those
lesser
information
to
a
slice
explicit
to
the
user.
C
C
B
That
still
a
particular
law
exam
it
is
that
still
the
best
way
to
have
to
do
it,
I
mean
there's
no
reason
why
you're
kind
of
a
record
of
the
files
youthful
in
the
life
of
of
it,
inter
so
this
is
the
list
of
strength
or
whatever,
but
is
it
really
necessary
to
have?
To
put
you
know,
two
steps
and
depending
on
again
I
mean
it
also
depends
on
the.
B
The
tools
that
you
have
to
compile
those
that
list
of
files.
Finally,
from
your
own
numpy
data
directory,
that's
a
bit
of
a
pain
if
you
want
to
look
a
long
time
clear
it,
whereas
if
you
put
it
in
time
range,
you
know
these
the
object
or
maybe
some
wrapper
around
the
object.
If
you
want
to
have
those
two
like
the
factory
class
and
the
downloading
separate,
you
have
the
rapper
to
film
behind
those
two-
and
you
say:
I
just
want
from
you
know
this
this
time
range,
but
that's
actually
what
you
want.
C
A
A
C
A
A
If
you
want
that
functionality,
if
it
tracking
your
local
files
on
disk,
you
have
to
use
the
database
is
there's
no
major
overhead
to
the
date
ways.
It's
one
extra
Python
package
to
install
and
it
all
works,
probably
and
or
we
can
make
it
work.
It's
not.
It's
not
like
we're
asking
users
to
learn
how
to
write
sequel
queries
before
they
request
that
they
are.
You
know.
D
A
To
some
extent,
okay
well,
but
my
point
is
that
if
you're,
if
you
want
a
sample
objector,
will
automatically
know
what
files
you
have
available
to
you
and
create
objects
for
you
like
Danny's,
requesting
the
functionality
we
have
in
some
way
to
do.
That
is
the
database
model.
So
therefore,
we
should
use
that.
C
D
A
C
B
No
I
haven't
but
way
here
was
this
girl.
I
went
to
me
we're
talking
last
week.
I
was
like
I
could,
like
a
will,
be
really
helpful.
So
in
every
through
such
a,
they
did
there's
an
SS
w
database
ready
so
I
mean
like.
Is
it
conceptually
that
different
from
from
that?
If
you,
if
you
had,
if
you
want
to
construct
a
ghost
light
curve
through
ssw
I,
do
thank
you
the
object.
Go
you
give
the
doctor
a
time
range.
B
A
A
C
B
Fine,
but
my
point
purely
was
that
there
is
a
precedent
by
having
you
know
like
an
putting
in
a
time
range
in
an
object
for
telling
you
that
life
of
by
going
to
a
database
where
data
it.
It's
not
a
total
departure
from
anything
and
sounds
like
to
me.
It's
a
good
idea
in
at
least
idea
in
principle.
Yes
angle,
it
is
or
why
shouldn't
we
extend
that
to
make
it
stand
it.
If
we
can,
you
lie,
keep
track
of,
and
certainly
also.
A
A
tech
we
are
building,
we
have
built
or
our
building
the
unified,
downloaded
thing
new
dog,
and
it
has
by
design
the
same
interface
as
the
database.
So
if,
once
the
user
knows
how
to
use
one,
they
can
use
the
other.
C
A
I'm
just
saying
I
did
when
I
said
you
should
have
to
use
it.
I
don't
mean
you
should
have
to
use
it.
Like
always
I
just
mean.
If
you
want
that
functionality,
then
the
piece
of
software
we
have
the
provider
is
the
database
yeah,
you
sure
the
database,
because
you
know
that
even
provides
it.
That's
like
saying
I
will
took
you
an
image,
but
I
don't
want
to
use
map.
That's
just
stupid.
B
You
make
a
life
of
a
big
like
a
practical
laughs,
then
adding
this
combined,
unified,
downloader
and
database
functionality,
possibly
with
a
rapper,
possibly
not
whatever,
but
if
we
allow
it
all
work
in
this
way,
it
doesn't
mean
that
you
can't
take
a
file
rename
it.
You
know
Jax
favorite
file
and
put
it
in
your
desktop
and
feed
that
fall
straight
into
the
isolated
light.
Curve
object
right,
at
least
in
the
odds,
isolated
factory
class.
A
Yes,
okay,
so
I
think
I
think
we
have
has
usual
gone
round
in
circles
for
a
while
and
come
back
to
this
place
where
we
started,
and
I
don't
think
they're
really
spectacular
disagreements
to
doing
what
we
have
planned,
which
is
the
like.
A
factory
class
out
on
file
names
explicitly
I
mean
I,
know
Dan
thinks
that
it
shouldn't
stay
like
that,
because
there
should
be
an
extra
wrapper
on
top
of
it.
A
Somehow
and
after
our
conversation
last
week,
I
have
a
design
for
this
floating
around
in
my
head
hope
you
write
up
a
separate
as
you
p,
for
it
as
I
think
it
will
make
quite
a
good
G
sub
project
with
you,
which
basically
involves
integrating
the
unified
download
into
the
database
and
changing
the
way
we
do.
The
current
search.
Caching,
because
the
database
of
the
moment
will
cache
of
ESO
queries
and
provide
you
the
data
back.
A
C
A
A
A
I
know
the
records
I
need
to
return
to
you
yet
have
these
records
instead
of
results
from
the
query,
and
it
would
think
but
I
don't
know,
is
that
if
we
sketch
results
from
the
query
instead
of
the
actual
query,
then
that
allows
us
more
fine-grained
control
over.
There
is
lot
results
of
the
query.
We
need
to
be
fetch
the
data
for,
but
that
is
something
I
need
to
work
on.
I,
don't
really
want
to
go
into
it
alone,
but
basically
it.
B
B
Be
like
that
without
would
be
useful,
is
if
you
ask
for
no
month's
worth
of
data,
and
you
see
something
in
at
the
end,
and
you
want
an
expert
day
or,
for
example,
look
I
looked
at
January,
but
don't
look
now.
I
want
to
actually
do
some
analysis
on
Jan
in
February,
together,
combined,
for
instance,
affecting
certain
events.
B
A
Hmmm
yeah
anyway,
so
we
come
up
with
way
of
doing
that,
basically,
which
is
like
fine
grain,
cashing
their
database,
which
is
the
feature
we
really
wanted
out
the
database.
But
we
didn't
have
fun
to
develop
last
time
and
then
we
and
then
potentially
we
can
add
the
ability
to
throw
a
database
query
using
like
the
attribute's
syntax,
that
we
have
at
the
moment
into
a
factory
class
and
it
will
return
you
the
object,
having
gotten
the
data
from
either
local
or
remote,
depending
on
where
these
come
from.
A
A
B
You
could
have
function,
Elliott
Matt,
we're
now.
If
you
hear
you,
we
can
do
a
search
of
the
database
or
good
unified
download.
You
know,
however,
whatever
interface
you
want
initially
goes
now
is
that
you
could
have
some
be
word
like
you
know,
force,
download
or
replace,
or
something
where,
if
you
happen
to
know,
there's
new
data
out
there.
You
know,
then
you
can
force
it
to
repopulate
you
know
or
to
replace
your
old
files.
If
you
wanted
to
yeah.
A
So
yeah,
we
would
need
some
way
because
the
because
of
the
limitations
of
the
server
services
that
are
available,
we
would
need
some
way
for
the
user
to
go
redownload
with
its
data.
This
data
is
your.
These
records
in
the
database
are
now
shot,
get
me
new
ones
and
I,
don't
think
we
have.
That
shall
add
it
to
the
list
with
it.
C
A
B
And
it
did
have
some
guy,
like
you
know
no
date
downloaded
or
date
modified
or
is
that
does
actively
change
it?
If
you
download
overcome
two
completely
separate
copy,
the
file
yeah.
A
I
don't
know
I'm.
This
is
a
really
sticky
subject
and
I.
Don't
think
there
is
any
nice
way
of
implementing
it,
just
client-side
we
to
do
it
well,
we
would
need
cooperation
from
services
like
vso
and
probably
the
data
providers
that,
via
so
useless
as
well,
I've,
had
conversations
like
this
with
Joey
in
the
IRC
channel
before
and
not
and
not
got
anywhere.
A
A
Just
got
kicked
out
of
the
room.
I
was
hiding
in
ok
M,
the
yeah.
We,
you
could
have
some
method
on
the
database,
where
you
throw
a
query
to
the
method
of
the
database
and
all
the
results
that
that
query
finds
are
then
deleted
or
something
for
flagged
for
removal
normal.
What
replacement?
Anyway?
That's
fine
I'll.
Add
that
that's
a
good,
that's
a
good
thing.
I
did
not
have
it
on
the
list
of
features
that
would
be
needed
so
over
to
the
list.
A
A
C
A
That
is
something
else,
I'm
still
I,
don't
know,
I'm
still
alert
to
be
convinced
by
this.
As
a
concept
I
know
my
site,
our
cuzick
I
am
yet
to
be
provided
with
a
work
with
a
working
design.
That
I
think
will
actually
solve
any
problems,
because
currently
we
currently,
we
standardize
the
metadata
there
like
that
is.
Basically,
the
functionality
of
map
is
to
standardize
the
imaging
metadata.
That
is
what
it
does
and
I'd
not
sure
whether
there's
any
advantages
to
having
it
done
differently
that
the
real
I
don't
know
I
agree.
C
C
B
A
C
Suppose
we
really
should
have
like
a
better
use
kiss
for
the
well.
That
I
mean
I
can
see
something
for
like
a
region
of
interest.
Object,
know
like
the
time
range
would
be
useful,
would
be
part
of
a
region
of
them.
Chest
object
for
I
like
curve,
and
that
may
have
something
to
do.
You
know
you
can
imagine
similar
kind
of
thing
for
map
cube
door
mat
by
itself.
Some
can
opt
love
interest
divergent
a
little
bit
lat
yeah.
A
A
A
A
common
in
the
process
of
implementing
the
like
a
factory
we
will
have
to
do
something
with
like
a
metadata
yeah,
but
I,
don't
know
what
until
we
get
there
I
in
my
previous
experimentations,
with
like
a
factory
classes,
I
remember
managing
to
hack
around
this
problem.
Basically,
it
wasn't
like
a
meditative,
no.
D
A
Only
or
both
certainly
isn't
there
is
get
a
sack
and
everything
is
wrong
with
like
a
metadata
in
that
they're
either.
Isn't
any
it's
a
complete
shambles
like
it's
different.
Some
I
mean
at
least
with
map
Alicia,
always
getting
a
consistent
format
in
that
you're
always
getting
a
dictionary,
because
you're
invariably
we're
reading
a
fits
file.
We
like
a
metadata,
you
might
even
have
any
metadata
if
you're
reading
a
CSV
file,
if
you'll
feel
some
some
clipping
like
curves,
read
multiple
HD
use
out
of
Google.
How
many?
What.
D
C
B
C
A
There's
a
specific
implementation
detail.
The
point
where
this
becomes
tricky
is
when
you
are
concatenated,
multiple
likers
together.
So
if
you
take
to
like
of
objects
with
some
metadata
dictionary,
which
independently
is
fine,
if
you
then
add
them
together,
I
extend
the
time
range
one
with
the
other,
so
you've
got
two
goals
like
herbs
and
you
want
to
make
one
goes
like
because
they
happen
to
follow
along
from
each
other.
How
do
you,
apart
from
now,
you've
now
got
to
metadata
dictionaries
and
without
some
kind
of
standardization?
D
A
Never
really
got
anywhere
because
all
the
more
the
problem,
but
also
go
ahead
too,
if
you
had
two
instruments
together,
if
you
had
to
like
herbs
together,
you
this
simplest
way
of
men
of
keeping
the
metadata
is
just
to
keep
both
dictionaries
right.
You
had
to
light
goes
together.
You
add
one
pandas
time
flame
sign
frame
to
data
frame
to
another
pandas
dataframe.
You
have
one
data
frame,
that's
easy!
If
any
of
the
times
x
overlap,
if
any
of
the
rows
overlap
of
each
other,
pandas
will
spew
it
back
at
you
and
go.
A
You
can't
add
these
two
together,
stop
being
such
a
and
but
what
do
you
do
with
the
head
of
what
you
do
with
that
metadata
at
the
dictionary?
Because,
if
you
just
you
can
just
keep
both,
but
if
you're
adding
two
of
the
same
instruments
together,
you
now
have
a
load
of
redundant
data
and
the
validation,
the
invalidation
for
adding
the
two
together
and
needs
to
take
into
account.
The
metadata,
like
you
shouldn't,
add
a
ghost
time
series
in
one
wavelength
to
a
coach
times.
D
And
you
can
also
lose
the
link
between
the
data
and
appropriate
header
like
if
you
merged
five
files
together,
it's
no
longer
so
obvious
which
data
came
from
which
file
and
you
have
to
go
and
then
try
and
figure
that
out
again
so
they're
all
sorts
of
issues
there.
Let's
try
to
understand
the
data
from
the
headers
once
you've
merged
things.
D
D
B
That
is
obviously
it
you'd
obviously
have
a
solution
to
that.
So
you
could,
you
could
combine
all
the
dayglo
ways,
but
how
much
of
that
is
a
little?
How
much
is
there
any
of
that,
which
is
a
bit
above
beyond
hope?
It's
not
I
mean,
for
instance,
should
the
fact
that
that
is
very
difficult.
Stop
us
implementing
functionality
that
an
individual
scientists
doing
it
manually.
They
still
have
that
same
problem
anyway,
yeah.
A
B
So
so
the
fact
that
we
can't
come
up
with
the
perfect
solution,
which
would
be,
we
should
be
careful
that
the
facts
that
we
can't
come
with
the
focus
ocean
or
a
perfect
solution
is
very
difficult.
We're
going
to
be
careful
that
doesn't
stop
us
actually
improving
functionality
and
doing
the
things
that
can
be
done.
B
D
I
think
we
need
at
least
a
basic
solution,
so
we
can
present
this
I
mean
I
like
Stewart
said.
The
most
easy
thing
to
do
is
to
just
keep
all
the
methods
right,
it's
kind
of
ugly,
it's
very
ugly,
but
it
can
do
it
that
way,
but
it'd
be
nice
to
have
like
a
an
uber
meta.
If
you
like
so
summarizes,
summarizes.
You
know
one
method
that
summarizes
what's
in
your
data,
but
that's
harder
if
I
agree.
A
That
is
not
that
these
mobloc
girl,
or
that
chief
doing
it
in
the
first
place,
oki
first
pretty
afterwards.
You
know
yeah,
yes,
well,
depends
how
ugly
I
guess
it's
no
more
ugly
than
what
we've
got
on
the
moment,
which
is
nothing
we
can't
add
like
girls
together
our
solution
to
this
at
the
moment.
He's
don't
do
it,
none
of
it
exactly.
B
You
know,
but
also
it's
I
mean
what
we
have
is
a
walk
back.
Just
keeping
all
the
meta
data
is
in
like
a
list
of
dictionaries
of
whatever
is
no,
it's
no
ugly
ER
than
you
know
what
an
individual
do
on
a
case-by-case
basis,
their
own
stuff.
So
it's
I
mean
that
is
certainly
not
a
like
a
dis
improvement.
If
that's
even
a
word
from
what
you'd
be
doing
anyway,
yeah.
A
A
A
A
So
this
brings
us
back
to
the
fact
that
we
need
to
get
the
ficking
Phi.
Doke
p
are
merged.
Already
then
said
that
he's
going
to
start
working
on
the
actual
factory
implementation
which
can
be
done
independently
from
fido,
but
we
can't
have
one
in
master
without
the
other,
but
no,
we
can't
have
a
like
a
factory
in
photo
with
master
without
fido
in
master,
because
if
you
merge
like
a
factory,
then
you
delete
the
ability
to
download
files
you
like
of-
and
if
you
haven't,
got
something
else
in
to
replace
that
functionality.
C
A
They're
real,
currently
a
hunk
of
failing
tests.
We
do
have
some
new
contributor
in
the
IRC
channel,
who
is
between
revising
for
exams
fixing
slowly
hopefully,
but
if
anyone
wants
to
fix
some
tests,
there
is
also
got
to
do.
There
is
also
a
to-do
list
on
PR
1300,
but
mostly
that
can
be
summed
up
in
fix
the
test
and
write
the
document.