►
Description
In this public meeting, we had a talk from our friend Will Cohen, an urban planner from Boston.
Will is the maintainer of the Geo library and the author of Ovid and Aurelius.
https://github.com/factual/geo
https://github.com/willcohen/ovid
https://github.com/willcohen/aurelius
The talk began with a general introduction to GIS and geospatial analysis, mainly in the context of urban planning. Then, Will went through the history, the state of the art, and the vision for the future of GIS in Java, Clojure, and Clojurescript.
In the discussion, several members shared their problems and practices. We also discussed the possible next steps in building the Clojure GIS stack, and possible ways to collaborate.
Moderator: Zachary Teo
Text chat:
https://tinyurl.com/y4w2ff22
A
A
So
some
quick
introductions
of
myself,
my
name,
is
zachary.
I'm
an
aspiring,
clojure
developer
in
my
final
year
of
uni
in
singapore,
looking
for
a
job
enclosure,
hopefully
next
year,
yeah
and
just
quickly.
My
exposure
to
gis
is
with
regards
to
like
my
internship
from
last
year,
where
I
was
working
in
singapore's
urban
redevelopment,
basically
singapore's
land
panning
authority,
and
they
do
they
have
this
internal
gis
visualization
platform
for
their
containers.
A
A
So
just
to
let
anyone
know
in
case
for
when
you
ask
questions,
it
will
also
be
recorded
and
otherwise
I'll
invite
I'd
like
to
invite
everyone
to
turn
on
your
cameras.
I
think
most
of
you
all
have
done
so
already,
but
I
think
it
also
helps
the
definitely
helps
will
to
sort
of
gauge
whether
or
not
any
of
you
are
lost
or
particularly
interested
in
something.
So
you
can
stop
yeah,
and
if
you
have
any
questions,
I
think
we
can
leave
the
question
we'll
leave.
A
We'll
do
the
presentation
first
and
at
the
very
end,
then
we'll
have
like
a
period
of
discussion.
But
if
you
have
any
like
very
pressing
questions
that
require,
like
immediate
clarification,
feel
free
to
interrupt
or
you
can
send
a
message
and
I
might
find
a
better
opportunity
to
interrupt
will
and
later
on.
A
I
can
also
help
to
read
any
questions
if,
if
you
all
like
me
to
do
so
so
just
some
zoom
etiquette
just
for
those
who
are
here
if
you're,
not
speaking
I'll
politely,
ask
y'all
to
mute
your
mics.
It's
like
the
button
in
the
lower
left
corner
of
the
zoom
window,
yep
and
yeah
so
sort
of
mentioning
just
now
for
today's
agenda.
A
We
will
be
talking
about
the
state
of
the
geospatial
ecosystem,
foreclosure
and
java
in
2020,
and
that
will
be
primarily
based
on
will's
rhythm
and
and
what
he
would
and
once
willow
is
done
with
that
and
what
he
would
like
to
cover.
Then
we'll
enter
a
period
of
open
discussion
and
okay.
So
that's
all
from
me!
So
will
are
you
ready
to
begin.
B
B
So
I
suspect
that
there'll
be
plenty
of
things
to
discuss.
After
what
we
talk
about.
So
just
a
little
bit
about
me.
You
know
some
of
the
questions
that
were
posed
were
not
just
about
the
details
of
specific
things
related
to.
You
know
closure
programming,
but
sometimes
they
were
about
what
about
geospatial
analysis
in
general
or
what
even
about
urban
planning
or
the
connection
between
these
kinds
of
disciplines.
B
So
I
thought
it
might
be
useful
to
give
a
little
bit
of
context
about
how
I've
come
into
this
whole
crazy
thing
in
the
first
place,
so
I
am
an
urban
planner.
I
live
in
the
city
of
boston
in
the
united
states,
and
I've
been
working
on
projects
related
to
urban
planning
there.
Basically
ever
since
I
finished
graduate
school,
so
you
know
right
out
of
school.
B
I
made
a
lot
of
maps,
but
I
became
increasingly
interested
in
data
analysis
and
software
programming,
and
so
for
a
little
while
I
left
that
company
I've
co-founded
a
company
called
supernormal,
which
was
another
design
firm
in
the
area
still
in
the
boston
area,
but
there
there
was
a
much
more
clear
link
between
you
know
architectural
design
and
this
kind
of
world
of
geospatial
analysis,
and
that
was
where
I
found
myself
increasingly
drawn
to
closure,
which
I'll
get
into
a
little
bit
later.
You
know
the
short
story.
B
There,
though,
is
that
I
also
learned
that
the
process
of
running
a
company
is
a
difficult
thing,
and
it
also
clarified
for
me
that
I
actually
want
to
be
an
urban
planner
one
who
uses
you
know.
Software
and
data
analysis,
but
I
ended
up
coming
back
to
to
utl
and
I've.
For
the
last
couple
years,
I've
been
back
continuing
to
work
on
urban
planning
projects
in
the
same
kind
of
vein,
but
I'm
still
integrating.
You
know,
data
analysis
and
closure
in
day-to-day
work
slightly
separately.
B
For
the
last
you
know
three
years
or
so
I've
been
an
adjunct
faculty
member
at
boston
college
at
their
school
of
management,
their
business
school,
where
I
teach
a
course
on
gis,
it's
to
undergraduates
it's
to
business,
school
students,
it's
to
social
work,
students.
They
are
not
professional,
they
are
not
computer
programmers.
There
have
been
a
couple.
B
You
know
computer
science
majors,
but
what
it's
also
been
for
me
is
a
weekly
reminder
of
the
difficulties
of
geospatial
analysis
for
people
who
are
new
to
the
field
and
so
that
lens
of
trying
to
make
it
easier
for
for
new
people
has
been,
on
my
mind
very
much.
I
don't
know
that
the
tools
that
I've
been
creating
for
closure
just
yet
solve
all
of
that.
B
B
As
you
can
see,
all
the
people
around
me
are
both
new
englanders
with
their.
You
know,
ll
bean
boots,
but
you
know
on
the
desks.
You
can
see
a
lot
of
paper
like
large,
drawn.
You
know
a
lot
of
drawings.
You
can
see
cork
boards
with
you
know
just
buildings
designed
pinned
up
all
around
the
office
right.
I
am
the
only
person
at
this
company
who
does
computer
programming
so
to
speak.
B
A
lot
of
other
people
do
very
complicated
things
with
computers,
there's
parametric
design
tools,
there's
all
kinds
of
3d
and
building
information-
modeling
stuff,
but
you
know
I'm
the
only
person
with
a
copy
of
emacs
and
closure
up
at
the
time.
Although
you
know
at
the
moment
you
know
in
2015,
you
know
in
this
photograph
I
was
actually
not
yet
introduced
to
closure.
So
I
was
deep
in
the
world
of
sql
and
post
gis,
which
I'll
get
it
which
I'll
get
into
there.
B
Since
then,
you
know
I've
moved
to
the
other
side
of
that
big
room,
although
now
I'm
of
course
working
from
home
right
here
since
I
live
in
the
united
states
and
things
are
complicated
but
a
little
bit
about
you
know
what
the
end
product
of
what
I
do
is,
and
I
think
this
might
be
useful.
It's
context
is
you
know,
I'm
not
making
websites
so
much
at
the
moment.
B
Actually,
the
big
bulk
of
what
I
do
is
long
processes
that
result
in
you
know
little
books
of
of
what
the
master
plan
for
a
community
kind
of
is.
You
know
right
off
the
bat
you
can
see.
Oh,
this
is
about
community
process.
There's
mayor's
offices,
there's
committees,
you
know,
there's
not
computer
programmers
right
off
the
bat,
but
there's
plenty
of
room
for
analysis
right.
So
the
first
step
in
a
lot
of
this
has
been
that
you
know
before
I
even
get
to
the
parts
of
analysis.
B
Is
I'm
literally
just
going
to
middle
school
cafeterias
where
we
have
public
meetings
in
the
evening
and
we
put
people
we
have?
People
put.
You
know
green
and
red
sticky,
dots
on
poster
boards
or
I
go
to
events,
and
we
have
a
bunch
of
foam
boards
and
I'm
pointing
at
graphs
and
charts
and
maps
to
people.
B
So
a
big
part
of
my
job
and
what
urban
planning
is
is
before
you
get
to
the
analysis
you
have
to
get
to
feedback
from
the
community
about
what's
actually
happening
in
a
place
right
information
that
you
know
will
inform
the
maps
and
analyses
that
you're
doing
this
is
you
know
a
plan.
The
things
I'm
going
to
show
you
for
the
next
couple.
Slides
are
the
big
plan
that
I
most
recently
worked
on.
B
Right,
there's
a
lot
of
maps
and
you
know
stuff
which
ends
up
being
sets
of
statistics
which
make
their
way
out
into
charts
and
graphs
and
stuff.
I
can
use
gi.
You
know
gis
software
to
make
a
lot
of
these
things.
I
can
use
excel
or
I
could
use
you
know
r
or
you
know,
to
varying
degrees,
the
closure
ecosystem
for
statistics,
but
a
big
piece
of
it
isn't
just
maps
a
lot
of
it
is
still
geospatial
right.
B
You
know
a
very
simple
thing
right
here
is
this
map
here
you
know
on
the
left
right.
There's
a
set
of
classically
end
uses
that
urban
planners
talk
about
all
the
time.
There's
residential,
there's
commercial,
there's,
industrial,
there's
the
different
general
types
of
properties
and
a
big
piece
of
understanding.
What's
going
on
in
a
community
in
a
city
in
a
neighborhood
is
just
that
distribution.
It
gets
a
little
bit
complicated,
there's
condominiums
and
other
kinds
of
legal
structures
that
lead
to
multiple
pieces
of
property
stacked
on
top
of
one
another.
B
This
particular
gets
a
little
bit
bigger,
there's
larger
kinds
of
data
sets
that
don't
do
not
yet
rise
to
the
scale
of
you
know:
big
data
they're,
still
small
data,
they're,
just
big
data
right
stuff,
that
is
at
the
region
of
multiple
states
in
the
united
states
stuff.
That
goes
back
over
time.
Things
that
you
know
range
in
the
scale
of
many
many
gigabytes
of
data,
and
in
that
sense
it
it.
B
So
sometimes
it's
economic,
sometimes
it's
regional,
but
it's
still
tied
to
telling
a
story
clearly
for
people
which,
finally,
the
way
you
know
it
all
comes
together
in
the
end,
is
you
come
up
with
a
plan
for
the
future,
which
is
really
what
my
job
is
always
about,
which
is
telling
a
story?
Here's
what's
going
on
here's
what
I've
heard
from
the
public?
B
Here's
all
the
kinds
of
analysis,
we've
done
to
figure
out,
what's
actually
happening
in
a
more
advanced
way
and
here's
what
we
think
the
community
should
do
next
or
here's
what
the
community,
through
a
more
complicated
outreach
process.
Kind
of
agrees,
is
a
good
idea.
In
this
case.
You
know
we
should
be
expanding
certain
industrial
areas
because
they're
doing
really
really
well,
which
is
a
an
idea
at
odds
with
a
lot
of
manufacturing
in
the
united
states.
B
There's
areas
where
we
should
be
looking
for
more
kinds
of
density
for
downtown
development
and
then
there's
targeted
places
where
we
could
do
you
know
residential
blah,
blah
blah
blah.
That's
where
architects
and
other
kinds
of
fields
come
into
play.
You
know
the
architects
and
designers
actually
design
the
stuff
in
3d.
They
might
be
using
parametric
tools,
they
might
not
be,
and
then
all
the
stuff
kind
of
comes
together.
The
graphs
and
maps
inform
the
design
which
we
then
use
to
make
more
graphs
and
maps
to
kind
of
prove
the
point.
B
B
My
lens
toward
this
is
very
very
much
the
use
of
spatial
analysis
as
an
exploratory
tool
right
I
would
be
happy,
and
I
think
it
would
be
a
very
productive
conversation
to
talk
at
length
about
sort
of
the
more
large-scale
data
set,
analyses
that
a
lot
of
you
might
be
exploring
enclosure,
and
I
think
that
a
lot
of
the
ways
that
I'm
exploring
geospatial
data
should
lend
themselves
very
very
well
to
that.
But
my
draw
to
lisp,
you
know
originally
to
learn
programming
and
to
learn.
B
B
I
really
really
don't
want
to
reinvent
the
wheel.
I
think
that's
also
something
which
I
like
about
the
closure
ecosystem.
It's
ethos
that
there's
a
lot
of
good
stuff
already,
whether
it's
the
entire
jvm
ecosystem,
whether
it's
all
of
javascript
or
you
know
increasingly
in
the
world
of
data
science,
whether
it's
r
or
python,
and
so
there's
a
lot
of
points
where
the
ideas
about
what
geospatial
analysis
should
be
a
lot
of
cases.
B
Maybe
closure
isn't
the
right
place
for
certain
pieces
of
it,
which
means
that
I
really
want
to
focus
on
where
it
does
something
better,
which
is,
I
think
something
we
could
be
talking
about.
You
know,
after
my
presentation
here,
there's
one
other
piece
I
just
want
to
give
as
a
quick
aside,
which
is
part
of
why.
Actually,
my
if
you've
seen
my
focus
on
the
geo
library
and
some
of
my
closure
library,
writing
has
taken
a
bit
of
a
dip
in
the
last
couple
months.
B
It's
because
you
know
you
can
even
see
from
the
slides.
I
last
presented
a
lot
of
my
job
used
to
be
going
places
in
new
england
in
the
united
states
talking
to
people
at
close
distance,
and
you
know
having
a
lot
of
in-person
meetings
right.
That
aspect
of
the
field
of
urban
planning
has
been
totally
upended
right
now
because
of
the
pandemic.
Getting
good
community
feedback
from
a
room
from
remote
or
via
zoom
is
a
lot
harder,
and
so
a
lot
of
my
my
focus.
B
The
last
couple
months
has
been
out
of
necessity,
less
just
the
standard
thing
that
I'm
used
to
going
to
meetings
talking
to
people
and
making
maps,
and
it's
been
much
more
focused
on.
Actually,
how
do
I
solicit
feedback,
or
how
does
a
community
have
one
of
these
processes
from
afar
or
remotely,
and
so
this
is
where
actually
I'd
like
to
point
out
that
the
closure
ecosystem
has
come
to
the
rescue
right,
there's
a
tool.
I
actually
learned
about
it
through
the
science
closure
community
through
the
data
science
chat
on
zulip.
B
You
know
christopher
small
who's,
given
a
presentation
in
the
past
to
all
of
you,
you
know
he
and
a
lot
of
other
people
have
been
working
on
a
tool
called
polis,
which
is
actually
a
combination
of
a
javascript
website.
But
you
know
the
nuts
and
bolts
of
analysis
are
in
closure
right
now
and
it's
a
way
of
kind
of
by
the
way
this
example,
these
screenshots
are
not
something
I've
done,
they're
a
screenshot
from
a
case
study
that
was
available
on
the
polis
website
in
rome.
B
But
you
know
it
is
a
tool
that
allows
the
communities
to
ask
questions
of
one
another,
come
to
consensus,
talk
about
things
and
there's
a
lot
of
very
fancy
geospatial
or
just
no,
not
geo,
not
spatial
fancy
analysis
happening
using
closure
tools
to
come
to
sort
of
this
consensus,
building
and
multi-dimensional
analysis.
It
lets
you
think
about
groups.
B
It
lets
you
have
kind
of
that
in-person
conversation,
but
it's
able
to
do
it
via
a
tool
and
also
use
statistics
to
you
know,
tell
the
story
a
little
bit
more
fully
so
for
whatever
it's
worth.
I'm.
Actually,
though,
my
role
the
last
couple
months
in
writing.
Geospatial
analysis
tools
has
dropped
a
little
bit.
B
It
will
be
a
lot
easier
for
me
to
sort
of
pick
up
my
involvement
again
with
the
you
know
actual
closure
programming.
Once
I
get
back
into
the
nuts
and
bolts
of
it,
but
for
whatever
it's
worth,
if
you
don't
know
about
this
tool,
I
can
I
would
urge
you
to
look
at
it.
I
find
it
absolutely
fascinating
and
I'm
going
to
be
using
it
in
my
job,
so
that's
cool
too.
B
Okay,
so
I
think
a
little
bit
of
framing
what
on
earth
geospatial
analysis
is
and
how
to
use
it
and
how
closure
fits
in.
It
might
be
worth
actually
just
doing
a
very
quick
overview
of
what
actually
is
involved
in
geospatial
analysis,
and
then
we
can
see
where
some
things
already
do
the
job
very,
very
well,
and
some
you
know
where
there's
room
for
for
improvement.
B
B
On
the
one
hand,
you've
got
data
right,
you've
got
points,
you've
got
lines,
you've
got
polygons
and
you've
got
all
the
stuff
that
goes
with
it,
like
that's,
really
actually,
just
gis
data
shapes
and
stuff
associated
with
the
shapes
it
comes
in
a
lot
of
different
forms.
It
could
be
raster
in
many
cases
it's
vector
right.
Those
are
the
the
inevitable
gis
shape
files
from
esri
that
you
know
date
back
to
the
early
1990s
and
are
truly
the
worst
to
work
with,
but
are
the
lingua
franca
the
word
document
of
gis?
B
There's
the
geo
json,
which
is
wonderful
for
you,
know
web
mapping,
but
is
a
ginormous
file
compared
to
a
shapefile
and
has
many
other,
you
know
downsides,
there's
a
lot
of
other.
You
know
sort
of
slightly
more
esoteric
data
set
types
that
exist,
but
one
way
or
another.
You
have
to
get
all
these
into
something
usable
and
at
the
moment
the
two
really
big
tools
that
that
planners
use
are.
You
know,
a
graphical
tool
to
make
maps
which
is
not
usually
web
mapping.
B
It's
usually
like
esri
arcmap
or,
in
my
case
qgis,
to
make
the
map
itself
or
maybe
you're
doing
more
analysis
on
the
data,
and
you
could
either
do
that
in
qgis.
You
know
using
a
lot
of
the
point-and-click
menus
and
there's
a
lot
of
sort
of
more
advanced
stuff.
If
you
want
to
get
into
scripting
in
python
that
you
can
do
in
qgis
or
you
can
just
bite
the
bullet
and
do
stuff
with
database
work,
you
can
do
it
in
post-gis,
and
that
is
an
extension
to
postgresql
that
makes
postgresql
aware
of
spatial
data.
B
So
you
know
from
there.
You
have
two
options.
You
can
either
make
the
map
you
can
make
it
really
fancy.
Looking
you
can,
you
know,
export
it
out
to
a
picture
which
is
like
the
findings
that
you
have,
or
you
can
do
more
analysis.
You
make
a
layer,
you
modify
a
layer,
it
gives
you
a
new
layer,
you
do
more
analysis
and
then
you
know
that's
another
data
set,
and
then
you
bring
it
back
in
you
put
it
back
in
huge.
Is
you
put
it
back
in
your
postgresql
database?
B
What
ends
up
happening
in
a
lot
of
cases
is
either
you
make
a
bajillion,
temporary
or
intermediate
layers
that
get
you
to
where
you
want
to
go,
and
they
all
take
up
a
lot
of
space
in
your
like
layers
window
in
arcmap
or
qjs,
you
delete
your
intermediate
ones
and
you
call
it
a
day
or
you
got
a
bunch
of
temporary
layers
in
postgresql
or
postgis.
You
know
you
find
the
final
one
you
want.
You
can
delete
your
like
little
schema
of
temporary
stuff
for
your
intermediate
tables.
B
The
reason
why,
in
many
cases
it
becomes
layer
based
is
a
lot
of
times
the
analyses
themselves
require
the
use
of
other
optimizations
to
make
analysis
faster
for
qgis.
You
know
the
process
of
making
a
new
layer
a
new
shape
file,
a
new
something
like
you
know.
Something
else
is
there's
a
level
of
indexing
that
that
qjs
and
arcmap
can
do
to
allow
you
to
perform
an
operation
across
the
entire
shape
file
across
the
entire
data
set.
That
is
a
little
bit
faster
or
in
post
gis
a
lot
of
times.
B
If
you
have
these
multiple
steps,
you
oftentimes
want
to
make.
What's
called
a
spatial
index,
it's
just
like
another
index
of
a
column
and
a
table
and
push
sql,
but
it
does
it
to
shapes.
It
puts
little
bounding
boxes
around
every
geometry
in
your
table
in
post
gis
and
allows
you
to
run
across.
You
know
massive
multi
gigabyte.
You
know
it
could
be
multi-terabyte
tables
and
only
apply
your
operation
to
the
geometries
that
you
wanted
to
apply
to.
This
is
very,
very
important.
B
If
you
don't
have
that
indexing
step,
your
operations,
your
spatial
operations,
can
slow
down.
You
know
astronomically,
and
so
in
many
cases
these
intermediate
layers
are
inevitable.
To
sort
of
make
your
workflow
happen
in
a
timely
manner,
once
you're
done,
you
can
either
have
your
maps
be
done.
You
can
put
them
together
in
presentations,
powerpoint
indesign,
you
know
maybe
some
kind
of
website
integrating
the
findings,
but
in
the
end,
it's
data
that
leads
to
maps
that
leads
to
information,
that's
kind
of
the
field.
I
generally
work
in.
B
B
Had
it's
primarily
not
a
lot
of
deep.
You
know
spatial
analysis.
They
had
one
old
copy
of
arcgis
from
like
10
years
earlier
on
one
guy's
computer-
and
you
know
when
I
was
the
new
employee
he
was
like.
Well,
you
can
make
the
maps
with
it,
but
like
it's
on
my
computer
and
I
like
it,
so
you
can
do
it
either
after
work
or
when
I'm
at
lunch.
B
Whatever's
really
best
for
you,
which
was
not
so
convenient,
so
I
became
highly
motivated
to
look
at
open
source
software,
and
now
qgis
can
do,
I
would
say,
99.9
of
what
arcgis
can
do.
It
also
happens
to
be
free,
it
works
on
mac
and
it
is
actively
developed,
and
you
know
for
those
of
you
who
are
not
sure
about
the
value
of
qjs
versus
javascript
mapping
or
something
else
like
I
am.
I
am
making
no
statement
about
the
the
morality
that
say
the
national
security
agency.
B
B
It
is
a
shockingly
powerful
tool
and
I
everyone
who
I
work
with
now
at
the
company,
where
I
work
all
the
stuff
which
you
know
normally,
we
thought
we
had
to
be
using
arcgis
for
things
which
cities
still
think
they
need
to
be
using
it
for
100
transition
to
qjs
and
post.js
with
no
issue
at
all.
So
you
know
in
terms
of
tools
and
things
that
are
good
to
use.
There
are
some
questions
about
different
kinds
of
analyses,
stuff
for
community
gardens.
B
How
to
think
about
doing
you
know
just
being
houses
in
a
basic
level.
Qjs
will
get
you
a
lot
of.
The
way
like
it
is
is
very,
very
good
and
people
are
committing
to
it.
It
seems
like
every
few
minutes
and
there's
a
solid
release
happening
every
four
months.
It
is
a
shining
star
of
you
know
the
open
source
movements,
so
all
of
us
being
equal
like
before
you
jump
into
the
world
of
closure
before
you
worry
about
needing
to
get
into
all
these
complicated
things
like
qjs
might
actually
do
everything
you
need.
B
B
It
takes
some
finesse
to
make
the
maps
pretty,
but
there's
a
lot
there.
So
just
a
pitch
right,
don't
reinvent
the
wheel.
That's
one
piece
and
for
whatever
it's
worth
even
universities
are
now
giving
formal
talks
at
open
source
conferences
that
you
know
qjs
does
all
it
needs
to.
You
can
leave
ezra
behind
so
again.
Another
appeal
to
authority,
just
this
is
a
quick
example.
This
is
a
slightly
old
screenshot
from
a
few
years
ago,
but
like
just
so,
you
guys
understand
you
put
layers
in
right.
B
This
is
a
layer
from
the
natural
earth
data
set,
which
is
just
every
state
in
province,
the
united
states.
You
add
it
in
it's
a
shape
file.
It
shows
you
the
map,
like
it
shows
you,
the
data
set
and
there's
a
lot
of
buttons,
so
this
might
already
get
you
guys
in
a
little
bit
of
the
world.
That's
not
what
you
want
it's
graphically
based!
It's!
They
all
the
icons
kind
of
look
the
same,
and
that
is
a
downside
right.
The
good
news
is,
it's
very,
very
powerful.
B
All
those
buttons
do
what
you
want
a
gis
program
to
do,
but
you
have
to
use
this
program
for
many
many
years
to
begin
to
feel
comfortable
with
using
it
right.
I
use
this.
I
use
qjs
every
single
day.
I
have
pretty
much
every
single
day
for
the
better
part
of
a
decade
and
I
still
discover
new
things
about
it
every
single
day,
so
you
should
be.
B
If
you
are
making
maps
and
it's
not
exclusively
for
the
web,
you
might
find
a
lot
of
paints
off
just
by
using
this
this
tool,
I
would
say-
and
it's
available
to
all
of
you
right
now-
I'm
also
happy
to
answer
any
questions
about.
You
know
how
these
things
get
integrated,
but
a
lot
of
it's
there
as
soon
as
you
start
using
it
a
lot,
though,
you
find
that
sometimes
things
get
hard.
Maybe
you
have
a
data
set
which
you
know
isn't
that
big
your
computer
is
pretty
powerful.
B
You
have
you
know
more
than
eight
gigabytes
of
ram
on
your
laptop.
It
shouldn't
take
all
night
to
do
these
kinds
of
analyses.
You
know
these
are
things
like
where
you
say
I
want
to
intersect
one
big
layer
of
every
piece
of
property
in
a
city
and
boston,
for
example.
Isn't
that
big,
a
city?
It's
like
half
a
million
people
plus
right.
B
You
want
to
intersect
every
single
parcel
with
open
space,
something
like
that.
That
seems
like
it
should
be
doable
in
a
fairly
short
order,
a
lot
of
times
doing
it
in
a
tool
like
qjs
or
arcmap.
You
know
your
computer
will
run
all
night
in
a
fairly
small
operation.
That's
a
good
sign
that
you
have
now
sort
of
reached
the
logical
limit
of
where
desktop
gis
programs
solve
your
problem
for
you
and
where
the
logical
next
step
is
to
jump
to
postgresql,
I'm
not
going
to
give
a
talk
on
sql
today.
B
If
you
want
to
learn
sql,
there
are
many
resources
available
for
you
all
over
the
internet.
The
jvm
is
fantastic
because
of
jdbc
like
if
nothing
else,
you
can
write.
Jdbc
statements
send
g
sql
queries
via
the
jvm
to
post
gis.
It
will
do
all
of
your
spatial
analysis
and
it
will
send
back
your
results
to
you.
It's
annoying.
You
know.
Maybe
you
know,
there's
a
lot
of
there's
a
lot
of
friction
involved
in
the
process
of
writing
of
going
from
java
or
javascript
one
site.
You
know
finish
a
port
of
the
javascript.
B
You
know
geo
library
out
to
you,
know
a
database
and
back
it's
not
always
the
most
preferable
thing,
but
in
situations
where
you
have
big
data
sets
not
so
big
that
they
turn
into
a
world
of
big
data,
but
somewhat
large
data
sets
that
you
want
to
be
doing
spatial
operations.
For
you
know
this
sql
statement,
kind
of
sums
up
the
power
of
what
postgis
is
right.
B
You've
got
two
tables
city
and
superhero
right
and
the
city
is
spatial.
There's
a
geometry
column
in
city
and
superheroes
are
spatial
for
some
reason.
I'm
not
quite
sure
why,
and
you
know,
if
you
want
to
figure
out
which
superhero
lives
in
which
city
you
can
use.
This
st
underscore
contains
function
in
the
where
clause
in
the
sql
statement
that
right
there
is
actually
post
gis.
In
a
nutshell,
it's
an
addition
to
postgresql
functions
that
start
with
st
underscore
can
do
spatial
things
to
geometry
columns
in
this
case.
B
If
the
city
contains
the
superhero,
if
the
superhero
is
fully
contained
in
the
city,
then
that's
a
true
one.
The
join
can
continue
and
you
can
provide
which
superhero
lives
in
that
city
right.
It's
just
like
doing
a
spatial
join,
you
know
in
qgis
or
arcmap,
and
you
get
back
a
a
a
in
this
case.
It's
saying,
select
superhero.name,
so
you
want
to
get
the
name,
but
you
could
say,
select
superhero.js
and
get
a
geometry
return
to
you
as
well.
B
So
you
know
qjs
plus,
post
gis
might
solve
all
your
problems,
and
all
you
might,
if
you
want
need
to
be
using
closure
for,
is
to
write
the
sql
statements
and
there's
a
lot
of
very
powerful.
You
know
libraries
available
in
closure
right
now
that
allow,
for
the
pair
parameterization
of
you,
know,
sql
statements
to
make
this
more
flexible.
B
Okay,
so
I've
already
started
talking
about
geometry
data.
So
then
that
comes
into
the
next
question.
What
what
is
gis
data?
How
does
it
show
up?
How
do
we
use
it
in
the
closure
world
it
just
I
I
this
sounds
very
basic,
but
it's
worth
stating
there's
really
a
couple.
Different
kinds
of
vector
spatial
data.
You've
got
the
simplest
thing,
a
point
right:
a
dot,
zero
dimensional
it
it
exists
somewhere,
then
you've
got
lines.
Lines
are
basically
dots
that
are
connected
to
each
other.
B
Right
multiple
points
connected
together
that
is
sort
of
one
dimensional.
Then
you
have
the
closed
version
of
a
line
where
the
lined
endpoint
is
exactly
where
it
starts
again.
That
makes
sort
of
what's
called
a
poly
line
or
a
multi-line
or
a
line
string.
You
know
in
the
world
of
gis
that
can
be
open
where
it's
just
you
know,
still
a
one-dimensional
shape,
but
it
becomes
kind
of
two-dimensional,
but
it's
just
the
boundary
of
it
or
it
could
be
closed.
B
It
can
be
a
polygon
or
a
multi
polygon
where
it
you
know,
has
an
area,
that's
that
and
then
sometimes
there's
a
world
of
making
it.
You
know
three-dimensional,
but
but
that
is
still
kind
of
a
not
fully
solved
problem
in
the
world
of
gis,
but
there
there
is
efforts
towards
that.
In
both
you
know
the
sort
of
sea
world
and
also
within
you
know,
post.js
to
think
about
that
stuff
and
then
there's
all
the
data
that
comes
with
it.
B
The
stuff
that's
associated
with
the
shapes,
this
blobby
shape
might
be
called
the
city
of
austin.
Maybe
it
has
a
population
of
650
000..
Maybe
it's
in
the
united
states
in
the
state
of
massachusetts,
whatever
so
you've
got,
shapes
you've
got
stuff
with
the
shapes
and
they're
somehow
connected
to
each
other
and
all
is
peachy
with
the
world
right.
Well,
no,
and
now
I'm
going
to
get
a
little
bit
metaphysical
for
you
all,
which
is
the
idea
of
coordinates.
The
shapes
have
to
exist
somewhere.
B
There's
nothing
inherently
earth-based
about
the
idea
of
shapes
right.
Actually,
the
way
js
tends
to
think
about
this
is
that
it's
cartesian
right.
They
exist
on
a
plane,
there's
an
origin
somewhere,
there's
positive,
there's
negative
quadrants
and
they're
just
there
on
it
right.
This
point
could
be
at
one
one:
the
polyline,
the
first
one
could
be.
You
know
the
first
one
could
be
a
two
zero.
B
B
All
the
way
down
to
the
south
pole
in
the
equator
longitude
tends
to
be
or
not,
intensity.
Longitude
is
the
vertical
lines
which
tell
you
east
and
west
off
of
you
know
the
prime
meridian
or
international
date
line
right,
and
you
know
just
one
quick
tip
for
anyone.
Who's
ever
started
to
deal
with
this.
You
think
about
xy
points,
and
then
you
think
about
latitude
longitude
points
and
a
very
common
mistake.
Is
you
flip
them
right
latitude?
B
If
you
think
latitude
longitude,
that's
actually
y
than
x,
the
lines
are
horizontal
in
latitude,
but
they're
talking
about
the
y
dimension
up
and
down,
and
so
even
if
you're,
just
thinking
in
latitude
longitude,
sometimes
you're
thinking
about
things
in
the
x
y
plane
other
times,
you're
thinking
about
things
in
the
latitude
longitude
field
and
there's
already.
This
idea
of
thinking
about
coordinates
relative
to
one
another.
B
That,
in
a
more
complicated
way,
is
the
world
of
the
projection,
which
is
really
getting
the
three-dimensional
earth
which
is
spherical-ish
onto
a
two-dimensional
plane,
and
so
the
best
way
to
think
about
it
is
the
idea
of
different
maps
that
we've
seen
right
back
in
the
day
back
when
europe
was
busy
drawing
lines
across
the
atlantic
ocean
to
conquer
the
new
world.
The
mercator
map
was
a
particularly
productive
map
projection
to
be
using
right
on
when
you
stretch
out
the
earth
this
way
to
get
it.
B
You
know
from
a
three-dimensional
ball
to
a
two-dimensional
plane
if
you
stretch
it
out
this
way.
The
mercator
projection
straight
lines
on
earth
are
straight
on
the
map.
This
is
really
really
good
for
cross
for
scaling
a
boat
across
the
ocean
right.
The
downside
to
this,
of
course,
is
that
size
becomes
drastically
distorted,
there's
always
a
give
and
a
take
the
catch.
The
projection
is,
you
can
have
certain
things
be
good,
but
other
things
get
distorted
as
you
go
from
three
dimensions
to
two
dimensions.
B
In
this
case
the
the
the
orientation
of
lines
stays
good
at
the
expense
of
other
things.
Maybe
you
don't
really
need
to
be
conquering
the
rest
of
the
world,
so
you
might
want
to
use
this
one
right.
You're,
gonna,
unpeel
the
earth
like
it's
a
globe
or
like
it's
an
orange
right.
It
all
stays
relatively
contiguous.
B
You
know.
Actually
this
would
be
a
very,
very
bad
map
for
crossing
the
atlantic
ocean,
but
what
happens
is
you're
able
to
both
preserve
area
very,
very
well,
or
you
know,
and
you're
generally
able
to
have
the
orientation
of
land
masses
be
good
for
one
another.
But
you
know
the
oceans
are
a
complete
mess,
so
a
different
kind
of
give
and
take-
or
maybe
you
know,
if
you're,
an
american
and
you're
used
to
reading
old
rand
mcnally,
atlases
you're
used
to
seeing
this
one
for
the
worldwide
map
the
robinson
projection.
B
B
There
are
these
ideas
of
conical
projections
where,
if
you
look
up
to
the
just
above
the
united
states,
there's
you
know
that
curved
latitude
line
and
then,
if
you
look
just
south
of
texas
and
florida,
there's
another
latitude
line.
The
idea
of
the
conical
projection
is
that
within
two
parallel
latitudes,
both
sides
and
shape
are
preserved,
but,
as
you
can
see,
they're
all
sort
of
the
longitude
lines
are
now
moving
toward
a
singularity
point
at
the
north,
and
so
this
would
be
a
very
bad
map
to
be
looking
at.
B
You
know
at
a
at
a
worldwide
kind
of
scale
that,
to
an
even
broader
level
brings
you
to
what
in
the
united
states
is
often
called
the
state
plane
projection,
there's
really
one
or
multiple
ones
of
these
for
every
state.
In
the
united
states,
most
other
countries
have
similar
divvying
up
of
their
countries
into
state
plane,
like
local
projections.
This
kind
of
solves
all
of
your
problems
at
a
local
level.
B
This
is
the
map
that
you
use.
If
you're,
making
a
map
within
the
state
of
massachusetts
north
is
up
area
is
generally
rectangular.
All
is
fine
right
if
you're
looking
in
massachusetts,
but
as
soon
as
you
zoom
out
the
world
falls
apart.
The
joke.
I
say
at
this
point
if
this
were
a
lecture
to
my
students
at
boston
colleges.
This
is
how
residents
of
massachusetts
look
at
the
rest
of
the
world
right
and
you
know,
but
the
the
key
here
is
that
data
comes
in
lots
of
different
projections.
B
The
reason
why,
if
I
go
back
to
this
previous
slide,
is
whoever
deserves
to
draw
the
poor
buildings
every
single
building
in
the
city
of
boston,
whatever
intern
at
the
you
know,
planning
department
had
to
draw
the
stuff
10
years
ago
and
is
keeping
it
all
up
to
date,
needs
to
be
using
a
state
plane
projection
to
do
it
at
a
high
level
of
accuracy,
and
the
same
is
true
for
every
other.
You
know
kind
of
city,
so
the
data
comes
in
a
bajillion,
different
projections
and
so
a
fundamental
role
of
any
good
gis
program.
B
Post
gis
qgis.
Whatever
else
is
to
be
able
to
navigate
between
all
these
things,
very,
very
well,
if
you're
just
doing
stuff
in
latitude
longitude,
this
sounds
like
a
lot
of
ruckus
for
no
good
reason
and
for
people
who
are
using
worldwide
or
just
point
based
data
in
latitude
or
longitude
projections.
Don't
really
come
up
just
as
a
quick
aside,
you
know,
project
map
projections
actually
even
make
their
way
into
xkcd.
B
This
is
yet
my
favorite
of
his
series
of
bad
map
projections,
the
oops
all
south
america
projection
set.
You
know
in
this
case
he
has
a
different
kind
of
compromise.
The
projection
does
a
good
job,
preserving
both
distance
and
azimuth
at
the
cost
of
exaggerating,
you
know
the
number
of
south
americans
so
there
it
is,
maybe
I'll
start
using
this
one.
In
my
analysis,
going
forward.
B
Okay,
fine,
a
lot
of
talk
about
projections,
where
what
is
the
point
right?
The
the
key
is
that
now
that
you
have
geometries
that
live
in
some
kind
of
projection,
gis
at
its
most
basic
level,
takes
different
shapes
that
are
on
this
cartesian
plane
in
some
form
of
projection
system
and
just
manipulates
them.
B
Maybe
it
literally
just
modifies
shapes
and
return
shapes,
so
here
right,
you've
got
a
and
b.
If
you
union,
a
and
b
together,
you
get,
you
know
a
blobby,
that's
a
and
b
together.
This
is
the
spatial
union.
Maybe
you
take
the
area,
that's
only
common
to
a
and
b
and
you
get
the
st
intersection
or
you
get
the
intersection
of
a
and
b
or
maybe
you
take
in
this
case.
You
know
liberty,
island.
This
would
be
in
new
york
city,
where
the
statue
of
liberty
is
and
you
could
buffer
out.
B
If
it
were
in
you
know,
the
state
plate.
The
new
york
state
meters
state
plane
projection.
You
could
buffer
out
500
units
from
this
thing,
which
is
already
in
a
meters
unit
system.
You
could
blow
it
out
500
meters
and
then
get
you
know
a
buffered
shape
for
those
of
you
who
are
looking
a
little
bit
more
closely
at
the
citation
for
where
this
diagram
comes
from.
This
is
post
gis
right,
st
union,
a
b
when
you
run
that
function
on
the
a
and
b
column
of
you
know.
B
In
this
case
it
would
be
a
single
feature.
A
single
row
of
shape
a
and
a
single
row
of
shape
b.
If
those
were
two
individual
tables,
and
you
did
st
union
on
them,
it
would
return.
It
would
spit
back
a
geometry,
a
geometry
object
in
postgas,
that
is,
that
shape
to
the
left,
or
you
know
the
intersection,
the
area
that's
common
to
the
two
and
that
uses
you
know
a
there's,
a
spatial
library
in
c
that
g
post
gis
runs
that
does
all
these
kinds
of
operations
right.
B
It
could
also
be
predicate
based
sort
of
boolean
kinds
of
operations.
Do
two
shapes
intersect.
Do
two
shapes
overlap.
Do
two
sheeps,
you
know
is
one
within
the
other
or
does
one
contain
the
other?
This
is
that
thing
about
gotham
and
the
superheroes
that
I
mentioned
at
the
beginning
right.
This
is
again
post
gis
right
you
can
intersect
overlap
within
contains.
Do
all
these
predicates
to
all
kinds
of
things
right.
This
is
another
thing
about
put
from
a
tutorial
on
post
gis.
B
You
can
see.
Does
a
point
intersect
with
the
line.
Does
a
point
intersect
with
the
polygon?
Does
a
multi-point
intersect
with
a
multi-line
string?
You
can
combine
all
the
different
geometries
to
one
another,
but
in
the
end,
gis
really
just
is
the
predicate
based
operations.
Do
certain
relationships
exist
and
when
you
take
two
shapes,
can
you
give
me
new
shapes?
That's
it.
If
you
learn
those
two
things
in
whatever
language,
that's
geospatial
analysis:
you
could
do
it
in
arcmap,
you
can
do
it
in
qjs.
B
You
can
do
it
in
post.js
and
you
can
do
it
in
closure.
So
the
lingua
franca
for
a
lot
of
tools
is
this
thing
called
geos.
It
is
the
engine
that
lives
underneath
post
gis,
it's
a
it's
a
c
c
plus
plus
whatever
it's
a
c
c,
plus
support.
It's
a
c
based
library
that
does
all
that
stuff
it
does.
It
does
intersections.
It
does
unions,
it
does
the
predicates.
It
understands
projections.
B
It's
it's
awesome
right,
it
is,
it
is,
is
robust.
It
is
mature.
It's
what
makes
post
gis
run
it's
what
makes
post.js
so
efficient
right.
It's
also,
if
you
do
gis
analysis
in
python
right.
It's
the
thing
that
lives
underneath
the
shapely
library,
which
is
really
the
thing
at
the
core
of
all
gis
analysis
of
python.
If
you
do
something
r,
it's
the
thing
that
lives
at
the
core
of
the
sf
librarian
r
right.
So
in
many
ways
it
seems
like
done
deal
just
use
that
right.
B
That
would
solve
all
your
problems,
but
we're
lucky
right.
Actually
geos,
which
is
the
lingua
franca,
is
itself
a
downstream
port
of
a
java
library
called
jts
right.
I'm
not
gonna
give
a
tutorial
of
jts
here,
but
in
a
weird
way.
What
I
will
just
say
is
at
a
very
basic
level
to
your
question,
to
the
questions
of
how
do
I
do
geospatial
analysis
and
closure?
The
answer
is,
it's
done
it's
solved
for
you,
sort
of
jts
is
in
java.
It
is
the
thing
which
actually
is
the
upstream
version
of
geos.
B
So
when
geos
gets
new
features,
they
actually
go
into
jts2
right
jts
has
been
around
for
a
long
long
time.
It's
been
recently
re-licensed.
It
used
to
be
lgpl.
Now
it's
eclipse
and
like
the
eclipse
distribution
library,
which
is
kind
of
like
bsd,
and
it's
it
does
everything
it
does
all
the
different
kinds
of
shapes
it
does
all
the
operations
you
make
java
objects,
these
gts
geometries.
They
could
be
points,
they
could
be
lines,
it
could
be
polygons,
and
then
you
can
do
all
these
things
to
it.
B
You
know
ancestry
a
lot
of
the
sort
of
from
a
couple
years
ago.
Heavy
hitters
in
the
world
of
closure
who
were
you
know,
did
different
at
different
times
stints
of
time
at
the
company
factual
that
is
sort
of
an
ad
technology
company
from
california.
A
lot
of
you
know
them
over
time,
put
together
this
closure
library
that
attempts
to
unify
the
jvm
to
digi
geospatial
analysis
when
I
first
started
learning
about
wanting
to
do
geospatial
analysis,
myself.
Basically,
I'd
kind
of
learned,
some
scheme
and
lisp.
B
So
when
I
first
came
to
it,
it
was
a
slightly
different
thing:
it
unified
three
geospatial,
not
five:
three
jvm
geospatial
libraries,
jts
spatial
4j
and
geohash-
that
seemed
very,
very
cool.
Actually
geohashes
are
a
neat
representation
that
kind
of
grids
up
the
world
into
a
set
of
recursive
little
rectangles
that
that
cover
the
globe
that
you
can
very
quickly
divvy
things
up
into
and
do
kind
of
analysis
right.
You
know
I
think
factual
uses
just
a
lot
to
kind
of
quickly
analyze
data
sets
at
sort
of
a
large
large
scale.
B
It
also
could
talk
to
spatial
4j,
which
is
a
neat
library
that
thinks
not
just
in
terms
of
these
cartesian
projections,
but
has
this
other
idea
which
is
sort
of
of
the
worldwide
projection?
What,
if
you
want
to
have
a
circle
right,
jts
is
point
based.
You
can't
actually
draw
a
circle
in
jts.
It's
not!
You
can
draw
a
thing
with
a
lot
of
dots,
that's
kind
of
like
a
circle,
but
not
a
pure
circle,
and
particularly
you
can't
draw
like
a
great
circle
that
circumnavigates
the
entire
globe,
because
jts
is
plane
based.
B
So
if
you
need
a
great
circle
or
if
you
need
something
that
thinks
about
the
poles
and
goes
across
at
all,
you
have
to
think
in
a
different
kind
of
way,
which
is
the
idea
that
spatial
4j
does
now
postjs
has
the
same
issue:
there's
geometries,
which
think
on
the
cartesian
plane
at
a
certain
projection
and
there's
geographies,
which
think
sort
of
globally,
and
so
the
jvm
version
of
that
is
gts
spatial
4g.
So
that
seemed
like
a
very
logical
place
to
start,
but
a
couple
different
things
came
up
for
me.
B
That
became
issues
first.
This
library
geo
did
not
do
anything
to
deal
with
projection
systems
which
made
a
lot
of
sense,
factual,
isn't
a
gis
programmer,
an
urban
planner
in
the
city
of
boston,
they're
doing
you
know
ad
and
data
technology
across
the
entire
globe.
Latitude
longitude
works
just
fine;
they
can
keep
it
in
the
same
projection
and
it's
never
an
issue.
But
a
lot
of
my
introduction
to
beginning
to
you
know
my
first
pull
requests.
B
My
first
contribution
to
the
library
became
around
adding
proj
4j,
which
is
the
pure
java
reimplantation
of
the
projection
modification
library
called
praj,
which
you
know,
exists
in
c.
It's
sort
of
the
java
implementation
of
it
it
it
can
reproject
points
from
one
projection
to
another,
and
you
know
with
geo
2.0,
basically,
proj
4j
got
added
into
geo
and
also
the
understanding
of
the
different
projection
systems
went
into
geo
as
well.
B
So,
in
addition
to
just
having
a
jts
object
in
jts
there's
this
field
for
the
the
which
projection
you're,
using
which
eps
g
code,
which
is
kind
of
the
open
source
numeric
system
that
standardizes
projections,
which
one
of
those
things
is
your
object
in
and
so
with
factual
or
with
geo
2.0.
You
had
a
project
4j,
you
had
a
jts
and
then
I
also
became
very
intrigued
by
this
idea
of
h3,
which
is
uber's
answer
to
the
geohash
right.
B
It's
a
hexagonal
indexing
system,
which
is
also
similarly
recursive,
but
has
a
lot
of
other
very
nice
features
for
urban
planners,
which
I
can
talk
about
at
length
some
other
time
so
that
was
sort
of
when
I
was
working
at
supernormal.
That
became
how
I
started
working
in
closure,
but
to
set
the
stage
for
that
sort
of
take
one,
my
first
three,
three
or
four
years
as
a
professional
urban
planner,
there
was
no
closure
at
all.
B
It
was
just
you
know,
making
a
lot
of
maps
and
doing
a
lot
of
post
gis
analysis,
and
this
is
kind
of
this
there's
no
map
here.
It's
actually
a
really
dumb,
looking
pie
chart,
but
it
was
the
pinnacle
of
my
gis
analysis
after
four
years
right
and
what
it
came
to
was
really
parcel
data
which
is
very
commonly
being
used
across
the
world
pieces
of
property
for
pieces
of
information.
B
B
So
actually,
if
you
just
purely
use
the
parcel
information,
you're
not
going
to
see
parks
or
roads
or
other
kinds
of
things,
it'll
be
called
government
owned,
but
we
wanted
to
have
one
that's
more
like
this
chart
to
the
right,
where
the
parks
and
the
roads
and
the
airport
and
all
the
other
kinds
of
pieces
would
be
pulled
out,
which
meant
that
it
was.
You
know
the
spatial
join.
B
Let's
take
all
the
different
layers,
all
the
different
multi-hundred
megabytes,
so
somewhat
small,
but
big
for
qgis
data
sets
and
do
a
lot
of
fairly
complicated
spatial
operations
on
it
clipping
out.
All
this
you
know
and
if
you've
ever
been
into
the
city
of
boston,
there's
a
lot
of
crazy
little
wiggly
parks
that
go
all
through
everything
like
very
carefully
extracting
out
each
piece
of
open
space
from
every
single
parcel
and
sort
of
operating
that
making
sure
you're.
You
know
doing
it
right.
B
So
that
means
first
unioning
every
single
piece
of
park
into
a
mega
park,
kind
of
an
uber
parcel
that
lives
across.
On
top
of
the
city
of
boston
and
then
individually,
differencing
that
out
from
every
single
parcel
in
the
city,
that's
a
fairly
simple
operation
but,
as
I
was
saying
before,
if
you're
not
making
all
those
fancy
indexes-
and
you
know
I
didn't
I'm
not
or
I
was
not,
especially
in
26
2015
2016.,
a
whiz
at
post.js
indexes
and
knowing
the
right
way
to
do
it.
B
You
know
I
would
write
notes
to
myself
pulling
out
all
the
open
space
like
it'll
work.
Just
be
aware
that
writing
that
sql
operation
is
going
to
take
me
like
an
hour,
and
actually
this
is
one
where
an
index
doesn't
even
help
you,
because
it's
just
you
have
to
take
a
really
complicated
space
and
then
individually
difference
it
on.
You
know
hundreds
of
thousands
of
parcels,
so
you
do
all
that
you
save
it
into
one
table.
B
Keep
that
like
spare
yourself
the
hour
going
forward,
and
then
you
know
over
the
course
of
a
week
or
two
talking
with
a
lot
of
my
coworkers.
I
did
kind
of
the
rebel
version
of
sql,
where
I
would
just
write
a
really
really
long,
sql
file,
and
just
one
one
by
one
run
a
bunch
of
commands
and
eventually
pull
out
all
the
different
pieces,
pull
out
all
the
open
space,
pull
out
all
the
airport
and
give
myself.
B
B
B
It
was
actually
a
question:
how
should
we
build
sidewalk
or
or
it's
a
cold
northern
american
city?
We
are
building
sidewalks.
You
know
we
have
sidewalks
every
winter.
I
think
there's
a
lot
of
snow,
we
plow
everything,
there's
potholes
the
stuff
gets
busted
all
the
time.
More
importantly,
in
the
winter
everyone
who
works
at
the
department
of
public
works,
you
know,
isn't
fixing
sidewalks,
there's
snow
everywhere.
You
can't
repair
roads
super
easily
in
the
winter.
So
a
lot
of
times
the
the
public
works
department
actually
have
their
employees
in
the
winter.
B
When
they
can't
be
sort
of
doing
all
their
major
capital
projects
go
and
ju
grade
the
condition
of
every
sidewalk
in
the
city
on
a
scale
from
zero
to
a
hundred.
So
they
know
how
broken
our
sidewalks
great.
So
then
the
question
becomes.
We
have
a
set
pot
of
money
and
city
government
to
fix
sidewalks,
which
one
should
we
fix?
We
can't
fix
them
all.
You
know
municipal
budgets
are
low,
so
the
task
that's
super
normal
was
to
do
an
analysis
for
the
city.
What
sidewalk
should
we
fix?
B
If
it
were
up
to
me,
will
the
urban
planner?
How
would
I
prioritize
that?
That's
a
classic,
geospatial
gis
analysis
question.
You
can
do
a
lot
of
different
pieces
right.
It
was
a
combination
of
a
lot
of
factors.
We
took
information
from
the
census
right
where
are
neighborhoods
that
have
historically
been
under
invested
in
or
the
city
has
public
policy
interest
in
trying
to
improve
the
you
know:
quality
of
life,
those
are
probably
neighborhoods
where
it'd
be
a
good
place
to
focus
and
there's
a
range
of
different
demographic
variables.
We
did
that
analysis
in.
B
Maybe
we
should
do
bang
for
the
buck.
We
should
fix
sidewalks
that
are
near
important
things,
so
we
found
a
set
of
assets,
schools,
hospitals,
public
transportation,
stops
parks,
all
kinds
of
other
things
did
walk,
shut
analyses
using
you
know
another
open
source
java
based
tool,
that
kind
of
does
isochrone
calculations
average
those
things
out
and
found
areas
that
are
close
to
assets,
the
city
of
boston
values,
then
there's
also,
which
sidewalks
are
the
most
broken.
B
Those
are
also
probably
good
sidewalks
to
fix
and
some
weighted
balance
between
all
that
stuff
can
lead
to.
You
know
this
was
the
final
product
right
cells
that
from
low
to
high
represented,
in
our
view,
a
balance
of
you
know,
a
combination
of
near
things
that
need
to
be
fixed
in
neighborhoods,
where
it'd
be
a
productive
thing
to
be
fixing
them
with.
Where
there's
assets
that
you
know,
people
walking
on
sidewalks
would
get
to,
and
this
maybe
the
exact
formula
didn't
make
their
way
into.
B
You
know
city
policy,
but
it
did
cause
them
to
rethink
their
their
process
for
giving
up
these
costs
and
how
they
thought
about.
You
know
fixing
sidewalks
in
the
future.
The
key
here,
though,
was
this
use
of
h3,
which
was
a
you
know.
It's
it
exists
in
c
exists
in
java
exists
in
javascript,
finding
a
way
to
actually
deploy
this
in
close
in
in
in
a
way
right,
and
so
there
that's,
where
closure
fit
in
very,
very
well
integrating
h3
into
the
geo
library
made
it
fairly
trivial
to
just
import.
B
You
know
the
city
of
boston,
shape,
poly,
fill
it
across
each
three
cells
at
a
certain
size
and
blow
the
thing
back
out
into
a
shape
file
now,
because
of
all
the
different
projection
issues.
Because
of
a
lot
of
the
other
issues
that
I'm
now
about
to
talk
about
that
still
are
pain
points.
A
lot
of
this
analysis
ended
up
being
in
post
gis
as
well,
which
is
totally
fine
right.
B
Post
js
is
a
very,
very
powerful
tool,
but
in
situations
where
it
falls
short
in
this
case,
h3
was
a
good
example,
because
it
wasn't
integrated.
Well,
closure
could
begin
to
fill
that
gap,
but
I
want
to
try
to
find
a
way
to
fill
that
gap
yet
again,
and
so
this
has
been
sort
of
my
kind
of
hobby
research
project
over
the
last
couple
years,
especially
as
I've
been
struggling
with
this.
You
know,
pain,
point
between
qgis
and
you
know,
post
gis
and
all
these
different
kinds
of
stuff.
B
So
I've
already
been
talking
about
you
know,
indices
right.
The
power
of
post
gis
in
particular,
is
the
ability
to
index
tables
and
sort
of
cache
a
lot
of
the
spatial
relationships
between
things
so
that
if
you're
gonna
do
intersections,
if
you're
gonna
check
for
things
it,
it
saves
you
time
whenever
possible
and
goes
to
kind
of
the
fastest
critical
path.
B
There
were
a
couple.
Actually
there
were
a
couple
different
talks
about
transducers
and
there'd
always
be
a
metaphor
like
sometimes
transducers
are
like
a
burrito
other
times.
It's
like
a
speaker
other
times.
It's
like
you
know,
airplane
baggage,
you
know
and
like
it,
you
know
that
that
was
hard
for
me
to
wrap
my
head
around,
but
eventually
it
kind
of
dawned
on
me
that
I
think
actually,
that
idea
lends
itself
very
very
well
to
for
me.
What
is
the
biggest
pain
point
in
gis?
B
B
Let's
say
we
want
to
get
the
intersection,
that
is
to
say
the
area
common
to
the
two
shapes
right
in
post,
gis
or
in
you
know,
various
versions
of
qjs
that
what
really
would
happen
is
this
dotted
line
would
come
into
play
and
the
question
is:
if
you,
if
it
knows,
you
have
two
shapes
two
circles
and
you
want
to
be
doing
a
spatial
operation
on
them
too.
You
want
to
get
the
intersection
there's.
Actually
three
tests
that
you
run.
B
First,
is
the
small
shape
the
blue
circle
inside
the
greens
or
the
gray
circle?
If
so,
if
it's
totally
inside,
then
actually
the
intersection's
a
really
fast
operation.
You
just
give
the
inside
circle
right.
If
the
small
one
is
totally
inside
the
big
one
and
properly
contained
there's
no
missing
holes
or
anything,
then
you
can
just
give
back
the
blue
shape.
You
don't
have
to
do
any
new
fancy
spatial
thing
if
they
partially
touch
right
one.
You
know
the
blue
kind
of
touches
the
gray,
but
is
not
totally
inside
or
totally
outside.
B
Then
you
have
to
do
the
complicated
thing
you
have
to
do
that
hard
spatial
intersection
operation.
Otherwise,
if
neither
of
those
things
is
true,
then
actually
there's
no
intersection.
It's
just
empty
right,
and
so
you
know
that's
kind
of
what
the
index
does
right
it
in
post
js.
It
will
figure
out
these
bounding
boxes.
These
dotted
lines.
B
It
will
find
the
relationship
between
the
two
shapes
and
do
its
best
to
optimize
to
the
fastest
path
drop
cases
where
there's
no
kind
of
intersects
or
no
kinds
of
contains
happening,
and
only
when
it
needs
to
will
it
do
these
more
complicated
operations.
So
that's
that's
the
key,
that's
what
we
need!
That's
what
that's
the
magic
of
post
gis,
sort
of
sort
of
the
indices
aside.
B
B
This
is
kind
of
my
first
attempt,
and
I
I
really
this
is
where
I've
been
sort
of
doing
a
lot
of
time,
to
think
to
make
sure
I'm
doing
this
the
right
way,
but
I've
been
putting
this
together
into
a
library
called
aurelius,
which
is
you
know.
I
was
reading
the
book
at
the
time
meditations,
and
these
are
my
meditations
on
doing
facial
analysis.
For
you
know,
for
as
an
urban
planner
right,
we
can
actually
write
out
an
intersection
function
that
wraps
jts.
B
B
Do
they
intercept
if
not,
but
they
still
intersect
in
some
way.
That's
the
second
level
of
the
conditional.
Then
we
have
to
do
the
hard
thing.
If
neither
of
those
two
things
are
true,
it's
just
it's
going
to
be
nil
right.
Those
are
the
three
steps
of
it.
That's
the
conditional
between
the
two
different
kinds
of
features.
B
You
can
kind
of
cue
the
thing
up.
You
can
take
the
big
open
space
layer
and
you
can
do
what's
called
a
prepared
geometry
in
gts.
This
is
a
feature
that
exists
in
gts
already,
where
it
kind
of
pre-bakes
these
bounding
boxes.
For
the
shape
so
that
you
can
very
quickly
do
a
contains
properly
or
an
intersex
test
on
operations,
so
you
can
feed
just
one
feature
into
the
intersection
function.
You
know
here
in
aurelius
and
it
will
prepare
one
of
the
two
geometries
and
set
up
a
transducer.
B
You
know
or
no
it
will
set
up
a
one
arity
function
which
can
go
into
a
transducer
into
map
into
filter
into
whatever
else.
In
other
words,
you
can
take
intersection.
You
can
make
all
the
other
jts
based
spatial
operations
and
using
map
or
filter
or
whatever
other
you
know,
closure
functions,
help
you
make
transducers.
You
can
apply
all
these
different
operations,
one
after
the
other
after
the
other,
and
it
will
take
advantage
of
all
this
prepared
geometry,
stuff,
preparing
and
unpreparing
as
is
needed.
B
What
aurelius
will
do
is
sort
of
put
all
the
things
together
to
try
to
get
you
to
the
optimal
path
for
operations,
so
that's
kind
of
where
I'm
working
right
now.
It
is
an
attempt
to
chain
the
different
things
next
to
one
another.
Now
I'm
I'm
I'm
I
my
my
primary
pain
point,
my
my
kryptonite.
In
a
superman
metaphor,
again,
once
again
is
documentation.
B
B
It
all
makes
sense
to
me,
because
it's
just
me
I,
as
far
as
I
know,
I'm
the
only
person
who
uses
this
library,
and
so
what
might
be
useful
is
it
is
to
create
more
of
a
conversation
around
the
closure
community,
about
where
this
makes
sense,
where
there's
flaws,
where
there's
pain
points
and
that
actually,
if,
if
anything,
I'm
not
I'm
not
trying
to
make
an
excuse,
I
do
need
to
write
more
documentation
for
it,
but
I'll,
I
think,
be
able
to
do
a
better
job.
Writing
introductory
documentation
for
it.
B
If
I
can
get
some
more
eyes
on
this
in
the
process,
but
anyway,
this
is
where
the
stands
right.
This
is
the
big
thing
I've
been
working
on
over
the
last
year
or
two,
an
attempt
to
take
all
the
operations
that
normally
I
just
have
to
put
in
post
gis
to
make
it
faster
but
prepare
myself
to
be
able
to
write
transducible
chains
to
do
it
in
in
aurelius.
B
So
over
the
last
couple
years,
this
actually
or
over
the
last
year,
really
since
I've
been
writing
it
more
recently,
this
is
actually
how
you've
been
doing
it
right.
So
you
can,
you
can
prepare
certain
features.
You
can
make
these
one
arity
functions
which
you
then
create
transducers
out
of
so,
for
example,
on
the
right
here.
B
This
is
an
example
of
an
analysis
that
I
was
doing
that
did
this
right
and
there's
no
map
actually,
because
no
map
was
needed.
This
was
a
question
related
to
that
master
plan.
From
before
that,
I
was
showing
you
pictures
of
a
thing
that
came
out
of
it
questions
about
the
rural
areas
of
the
city.
Maybe
we
should
be
thinking
about
densifying
them.
You
know,
of
course,
there's
questions
around
environmental
protection,
there's
all
kinds
of
other
policy
issues,
but
just
all
else
being
equal
were
we
to
change
zoning
and
change
the
rules
about.
B
You
know
development
in
certain
kinds
of
zones
in
the
city.
What
would
happen?
How
much
new
development
could
you
be
making
right?
And
that
was
where
you
know
a
a
transduceble
chain
come
into
play.
First,
we're
going
to
filter
some
stuff,
take
all
the
properties
pull
out
ones
that
are
into
a
certain
kind
of
zone.
Next,
let's
filter
it.
So
only
the
things
with
a
land
area
greater
than
blah
blah
blah
blah
still
are
there
right
to
meet
one
of
the
rules
of
zoning.
Next
comes
the
spatial
operations
right.
B
Let's
take
out
the
water,
because
the
analysis
can't
the
land
area
can
or
the
parcels
can
only
apply
the
areas
that
are
in
the
land.
So
I
would
write
a
little
function
like
difference
hydro,
that
was
basically
you
know
that
that
one
arity
function
where
I
prepared
the
water
layer-
and
I
said
if
you
ever
feed
this
thing-
a
parcel-
it
will
pull
out
the
water
and
it'll
return
you
the
water.
So
then,
if
you
run
map
difference,
water
it'll
take
out
the
waters
right
right
under
it
I'll
chain
it
to
the
next
thing.
B
Difference
wetlands
pull
out
wetlands
pull
out
open
space,
and
then
you
can
do
other
kinds
of
geospatial
analyses
that
you
map
one
after
the
other
after
the
other,
and
then
all
this
would
actually
do
in
the
end.
There's
no
maps,
it
was
just
I'm
sorry,
there's
no
visual
maps,
there's
plenty
of
closure
maps
right,
but
what
it
would
give
me
just
numbers
how
many
things
are
left,
how
much
areas
left,
and
that
was
the
answer
I
needed
just
how
much
area?
How
much
was
there
stuff?
Is
it
worth
doing?
B
B
One
of
those
really
long
sql
files
would
have
been
a
nightmare
like
it
would
be
out
more
thousands
of
lines
long,
because
every
one
of
these
steps
would
be
some
chain
thing
of
a
bajillion
st
functions
against
one
another,
but
something
about
the
functional
level
of
programming
having
these
operations
still
be
low
level
still
be
kind
of
tied
to
things
like
the
gi
or
the
gts
intersection,
but
being
able
to
chain
them
all
against
one.
Another
has
proved
for
me
at
least
very,
very
productive.
B
So
that's
that's
what
aurelius
does
for
me
to
make
that
happen.
It
actually
required
taking
a
previous
step
back,
which
is
another
library,
even
working
on
called
avid
you
can
see.
I
was
also
reading
some
greek
text
at
the
time
right.
This
is
metamorphosis
between
different
kinds
of
things,
so
the
power
of
the
geo
library
is
that
it
unifies.
You
know:
jts
spatial,
4j,
geohash,
now
proj4j
h3
into
sort
of
this.
B
This
one
protocol
based
abstraction,
called
the
shape
like
so
you
can
take
a
geometry
across
any
one
of
these
jvm
libraries
and
you
get
something
which
which
the
geo
library
can
just
operate
on
and
that's
great.
It
can
convert
to
gts
and
convert
it
to
shape
and
it
can
sort
of
think
of
it
conceptually.
So
anything
that
implements
shape
like
or
extent
you
know
does
shape,
like
extends
shape,
like
you
know,
can
get
all
these
things
as
well
right.
B
It's
just
a
new
protocol
called
feature
like
the
feature
is
the
gis
version
of
a
shape
and
its
property
and
its
attributes.
So
what
it
does
is
it
implements
a
couple
different.
You
know.
Well,
these
are
minus,
so
you
have
to
actually
use
the
function.
That's
without
the
minus
two
shape,
two
gts
geometry
properties,
and
then
you
can
change
both
the
geometries.
You
can
either
associate
a
new
one
or
update
it
with
a
function
or
do
the
same
thing
to
the
properties.
B
You
know
hash
map
field
and
the
point
there
is
you
put
in
a
feature
like
you
get
out
of
feature
like
it
can
be
just
a
shape
like
it
could
be
just
a
geometry:
you
can
do
your
operations
to
it
that
are
spatial
and
it'll.
Give
you
back
another
shape
like
you
could
give
it.
You
know
a
hash
map
with
both
the
geometry
and
the
properties
you
can
put
a
spatial
operation
onto
that
via
feature
like
and
it'll.
Give
you
back
a
modified
feature.
B
B
That's
what
aurelius
needs
to
do
its
work.
You
give
it
a
feature
like
contains.
Intersects
covers
intersection
union
and
instead
of
spitting
back
another
shape,
it
spits
back
a
feature
like
if
you
give
it
just
a
jts
object,
it
will
do
all
its
prepares
whatever
stuff
and
give
you
back.
Another
jts
object:
if
you
give
it,
you
know
a
closure
map
with
a
geometry
and
a
properties
field.
It
will
then
give
you
back.
You
know
a
modified
geometry
field
and
then
the
attributes
go
with
it.
B
This
means
that
just
like
qjs,
just
like
post
gis,
I
think
the
basic
tool
is
there
for
you
to
perform
these
change
of
operations,
think
about
it
spatially
or
think
about
it,
where
the
information
is
associated
along
with
it.
The
other
big
thing
is
the
other
main
pain
point
I
needed
to
sell
right
off.
The
bat
is
post.
Yes
itself.
If
you
talk
through
jdbc
and
you
ask
it
some
post
gis
query,
it
spits
back,
you
know
a
record
in
jdbc
and
you
get
some
funky.
B
You
know
you
know
binary-ish
kind
of
object
in
postgres
that
that
the
geometry
thing
represents.
So
the
other
big
thing
which
is
here
right
off
the
bat
is
extend
protocol
the
readable
column.
Basically,
if
the
object
in
one
of
the
fields
for
a
postgas
result
is
you
know
a
well-known
binary
hexadecimal,
which
is
the
way
that
sort
of
the
post
gis
jdbc
interoperability
works.
Then
it
will
read
it
in
natively
to
make
a
jts
object
using
the
geo.io.
B
You
know
name
set,
so
you
can
sort
of
up.
You
know
if
you
apply
the
the
db
namespace
into
your.
You
know,
jdbc
queries.
Basically,
it
will
modify
the
next
jdbc
connection
so
that,
if
you're
doing
postgas
back
and
forth
you'll
get
native
jts
objects.
Basically,
what
that
means
is
that
the
rows
that
we're
talking
about
kind
of
feel
like
features,
there's
a
little
bit
more
work.
I
need
to
do
to
kind
of
really
get
a
good
database
row
associated
with
a
feature
where
it
pulls
out
the
right
object.
B
B
The
interoperability
work
a
little
bit
better.
So
where,
where
am
I
going
with
all
this?
What
do
I
want
in
a
magical
world,
and
this
is
actually
not
closure
at
all?
This
is
something
that
you
know.
I
made
a
lot
of
maps
for,
but
one
of
my
co-workers
at
the
place
where
I
work
did
for
a
project
right
a
lot
of
times.
You'll
do
some
analysis.
You'll
make
a
bunch
of
maps.
B
You
know,
maybe
in
somerville
here's
a
vision
for
how
it
could
look,
but
the
community
process
needed
to
be
around
well,
I
don't
know
like.
Maybe
the
building
should
be
taller
or
shorter.
Maybe
there
should
be
a
different
mix
between
commercial
and
residential,
and
so
you
know
the
sort
of
best
tool
right
now
for
quick
drafting
of
three-dimensional
buildings
at
the
architectural
scale
you
know
for
cad
is
called
rhinoceros
rhino
and
it's
what
ever
it's
what
you
know.
Most
of
my
colleagues
are
drawing
out
their
sort
of
quick
schemes
for
architectural
design.
B
In
it's
great,
it's
a
very
powerful
tool
and
with
it
is
this
other
thing
called
grasshopper,
which
there's
a
screenshot
of
it
to
the
right
and
for
anyone
who,
like
knows
the
infamous
world
of
like
no
code
programming.
It's
like
the
holy
grail
thing,
which
is
never
fully
solved
right.
The
dream
that
we'll
be
able
to
make
some
like
you,
know,
weird,
connected
stuff
that
makes
people
not
to
write
code
anymore,
like
it
never
quite
makes
sense
right,
but
actually
for
grasshopper.
It
works.
B
B
You
have
to
do
a
lot
of
work
like
this
is
a
time
I
didn't
zoom
out
right,
but,
like
you
get
the
point
right,
there's
a
lot
of
things
that
are
all
connected
to
one
another
and
it
looks
like
a
tote
like
a
programmer
would
freak
out
at
this.
This
is
why
they
would
never
want
that's
the
way
they
write
code
right.
You
would
never
draw
some
crazy
mind
map
of
your
algorithm.
B
You
would
just
code
the
thing
out,
but
this
works
better
for
architects
if
they
think
more
visually
right,
but
you
know
I
look
at
this
thing
and
I
see
transducers
actually
like.
Actually,
this
is
the
thing.
This
is
the
idea
for
people
who
are
not.
You
know
necessarily
deep
closure.
Programmers
who
don't
want
to
learn
sql,
who
just
want
to
take
multiple
gis
based
steps
to
solve
their
question.
That's
that
to
me
is
the
dream
that
you
shouldn't
have
to
type
all
this
junk
out.
B
My
efforts
kind
of
over
the
last
year,
it's
been
a
sort
of
a
fundamental
piece
that
required
thinking
about
odd
and
aurelius,
although
right
now,
as
far
as
I
know,
I'm
still
really
the
only
kind
of
user.
So
I
haven't
even
released
an
initial
alpha
release
before
I
get
there,
there's
actually
another
thing
which
ended
up.
You
know
I
optimistically
thought
it
would
be
like
a
couple
months.
I
was
telling
daniel
earlier.
I
thought
it
could
be
done
in
august.
It's
now
october.
B
B
It
should
be
very
simple
for
me
to
just
translate
the
geo
library
from
clj
to
cljc,
just
replace
all
the
gts
interop
in
that
in
the
jts
namespace
of
geo
and
have
it
all
just
instantly
work
on
j
in
javascript
conceptually.
That
was
what
I
thought
a
few
months
ago.
I've
now
run
into
there's
a
lot
of
caveats.
There's
a
lot
of
small
differences
between
java
and
javascript.
B
You
know
advanced
compilation
is
an
essential
piece
of
any
good
library
that
has
been
like
my
I've,
been
trapped
in
the
world
of
advanced
compilation
for
the
last
couple
months
trying
to
get
to
work,
but
the
end
is
in
sight.
I
I
am
working
on
it.
I
I've
gotten
a
couple
different
name
spaces
going.
I
hope
soon.
Whatever
soon
is
to
get
that
done,
and
then
at
that
point
the
geo
library,
all
the
sort
of
idioms
of
doing
closure.
B
Spatial
programming
should
work
in
the
world
of
closure
script
as
well,
which
I
think
could
be
very,
very
powerful
that
you
can
just
take
the
objects
geojsons
whatever
and
in
the
same
way
you
can
just
do
all
the
stuff
in
jts.
You
can
do
it
in
jsts
and
geohashes,
and
you
know
h3
as
well,
so
that
that's
the
big
thing
I've
been
working
on
the
last
couple
months.
Once
I
get
that
completed,
I
want
to
just
update
audio
and
aurelius
to
do
the
same.
B
B
Sqlite
requires
in
java
native
libraries
to
operate,
so
this
might
be
an
opportunity
actually
to
just
sort
of
take
the
leap
and
finally
finish,
a
new
library
that
other
that
the
proj
project
has
been
working
on,
which
is
this
thing
called
proj
jni,
which
is
an
attempt
to
actually
codify
the
c
to
write
good
bindings
around
the
native
version
of
spatial
projections
and
and
write
a
java
library
that
works
around
that
there's
a
couple
different
steps.
It
needs
to
be
like
accepted
as
a
new
osgo
project.
B
If
it
happens
like
we
need
to
find
a
way
to
embed
all
the
different
native
platforms
into
a
jar,
so
it's
like
good
for
java
users,
so
I'm
kind
of
working
down
that
a
little
bit
too,
although
I
need
to
get
back
into
it
but
like.
I
think
that
that
is
a
really
powerful
step
for
the
jvm,
because
what
will
mean
is
that
the
parts
where
the
jvm
falls
short,
where
it
doesn't
understand
all
different
projection
systems,
where
there's
errors
where
you
have
to
start
doing
weird
interop
with
c.
That's
very,
very
painful.
B
I
think
if
the
projections
get
better
fixed
in
the
jvm,
then
it's
easier
to
think
that
many
of
the
operations
in
geo
in
in
gis
will
work
with
less
pain
on
the
jvm
and
then.
Finally,
the
other
big
task
that
I
know
needs
to
happen.
Is
there
just
needs
to
be
more
documentation
like
this
talk?
Is
a
good
start,
it's
kind
of
a
ramble,
there's,
not
a
lot
of
detail.
B
I
know
that
one
thing
I
need
to
do
is
actually
just
really
do
a
better
job
documenting
the
stuff,
how
I
use
it,
how
other
people
are
using
it
and
get
that
clear
so
that
it's
it's
less,
you
know
difficult
for
newcomers,
so
that's
sort
of
what
I
see
as
clear.
Next
steps,
based
on
some
of
the
questions
from
you
all.
I
see
there's
sort
of
other
directions
that
are
good
next
steps
too.
What
about
big
data?
I
haven't
talked
about
this
at
all.
B
There's
the
technl
data
set
library
right,
there's,
which
is
complicated
and
new
and
rapidly
changing,
there's
a
long
process
of
trying
to
figure
out
how
to
get
geospatial
stuff
into
that
right,
which
is
an
unsolved
problem.
What
about
stuff
that's
bigger
than
just
postgres
right?
There's!
You
know
sort
of
the
world
of
spark
a
lot
of
stuff
on
the
gvm
there
were.
You
know
some
comments
about.
You
know
geospark
now,
sedona
and
finding
ways
to
integrate
that
it
looks
to
be
a
very,
very,
very
powerful
tool.
B
I'm
probably
not
the
right
person
to
try
to
push
that
one
forward,
just
because
my
data
isn't
usually
big
enough,
but
I
would
love
to
figure
out
a
way
to
have
our
the
efforts
that
I'm
working
on
sort
of
make
that
easier.
So
I
need
some
guidance
there.
Then
there
are
also
questions
about
mapping.
Visualization
and
sort
of.
B
I
would
pose
it
back
as
questions
for
the
group
about
when
maybe
qjs
is
enough
or
when
you
really
do
need
a
web
map
in
terms
of
analysis,
it's
sort
of
like
I
was
just
saying:
when
is
post
gis
enough
when
is
jdbc
enough,
so
to
speak
or
when
would
something
like
transducers,
maybe
make
it
faster
or
if
you
have
to
jump
up
a
level
when
would
something
like
tech
ml
come
into
play
in
the
world
of
front
end?
There's
a
lot
of
different
pieces
right
at
the
moment.
B
A
B
Yes,
it's
great,
I
I
have
used
it
reasons
why
I
haven't
used
it.
Super
super
deeply
are
just
that
it's
kind
of
a
little
bit
annoying
to
like
stand
up
the
fairly
complicated
piece
of
software
entirely
separately,
as
it
becomes
its
own
kind
of
world
and
also
in
many
ways
I
haven't
found
it
to
be
better
than
that
combination
of
postgas
and
qjs
or
post
js
plus.
B
You
know
some
web
mapping
right
and
you
know
some
of
the
work
that
I
was
doing
at
supernormal
was
a
combination
of
these
kinds
of
analyses
that
would
lead
to
a
web
map
without
having
to
jump
all
the
way
to
cardo
and
doing
it
the
cardo
way
of
thinking
the
other.
I
guess
piece
is,
if
you
know
a
lot
of
the
analyses
that
I'm
doing
you
can
see,
it
was
like
I'm
literally
just
thinking
about
intersections
in
buffers
and
filtering
operations
like
cardo.
B
Does
those
things
but
part
of
cardo's
magic
is
doing
those
things
and
providing
a
clear,
user-friendly
tool
doing
it
over
the
web
right?
That
does
not
deal
with
projections,
particularly
well.
That
does
not
allow
for
the
level
of
styling
that
maybe
qis
can
do
as
well,
and
so
in
many
cases,
cardo
solves
a
different
problem
than
the
problems
that
I
have.
That
said,
it's
an
amazing
tool
and
you
know
if
you
find
that
it
solves
a
lot
of
your
pain
points.
B
C
C
B
Yeah,
I
think
it's
true,
but
it's,
but
it's
still
great,
like
it's,
it's
very,
very
useful
and
if
you're
trying
to
operate
stuff
across
a
lot
of
different
people
who
are
trying
to
work
together
on
data
sets,
if
you're
trying
to
make
something
where
you
know
the
back
end
is
the
part
that
you
want
to
handle,
and
you
want
to
simplify
a
web-based
interface
to
get
that
stuff
kind
of
done.
Then
yeah,
I
think
cardo
could
solve
your
problems
for
you.
I
see
a
second
question
about.
B
How
can
I
read
every
shape
files
and
use
them
in
geo?
You
have
the
the
biggest
problem
of
all
so
turns
out
shape
files
are
the
worst
I
mean
they're
great.
They
do
a
lot
of
things,
but
there's
a
lot
of
problems
with
them.
They're
hard
to
read,
there's
a
lot
of
weird
reverse
engineering.
You
know
geotools
and
a
lot
of
powerful
javascript
light.
You
know
jdm
libraries
handle
it,
but
they
kind
of
feel
like
bringing
a
sledgehammer
to
a
knife
fight
which
sounds
kind
of
violent,
but
I
think
you
get
my
point.
B
B
I
started
going
through
the
specification
for
shapefiles
and
from
a
lot
of
them
and
try
for
geom
basic
geometries
lines
and
points
and
stuff,
and
I
basically
did
begin
to
implement
in
jvm.
This
is
the
part
that
maybe
won't
directly
translate
to
javascript,
and
I
made
it
so
the
dot
shp
you
can
read
it
in
there's
a
lot
of
problems.
You
can
already
hear
it
right.
So
the
problem
with
the
shape
file
for
those
of
you
who
don't
use
a
lot
of
shape
files.
Is
it's
not
just
one
file.
B
It
actually
is
like
eight
files
that
are
all
next
to
each
other,
there's
the
dot
shp,
which
reads
the
shapes,
just
the
shapes
just
no
projection,
just
where
the
shapes
are.
There's
the
dot
dbf,
which
is
all
the
attributes,
which
is
an
old
database
format,
called
dbase
three,
which
is
kind
which,
like
you,
can
kind
of
open
in
excel
and
like
it's,
it's
the
sort
of
attribute
store,
there's
a
dot
prj,
which
is
a
it's
called
a
well-known
text.
B
So
when
I
find
myself,
what's
it
it's
on
my
list
once
I
finish
the
cljs
part,
one
section,
I
think
I
do
really
really
really
want
to
get
the
completion
of
shapehouse
support
if
you,
if
someone
else
wants
to
take
that
on
instead,
I
would
be
happy
to
accept
a
full
request,
but
I
do
recognize
this
key,
so
I
have
the
dot
p.
That's
I
have
sort
of
partial.
Yes,
that's
correct!
There's!
Actually
let
me
answer
it
two
different
ways:
one
you
can
read
the
shp
file.
B
You
can
then
deal
with
the
projection
yourself
and
assign
it
somehow
in
the
jvm.
You
can
read
the
dbf
using
other
kinds
of
java
libraries
and
you
can
make
all
this
stuff
together
now
an
easier
answer.
You
download
the
shapefile,
you
install
qjs
on
your
computer,
you
open
the
shapefile
in
qjs
and
you
save
it
as
a
geojson.
That's
actually
right
now
how
I
get
stuff
in.
I
either
read
in
tables
from
post
gis
or
just
at
the
moment.
B
Until
I
fix
all
the
shapefile
support,
I
just
use
qjs
to
do
the
conversion,
and
then
I
open
it
back
up
as
a
geojson.
The
question
was,
you
know:
if
it's
hard
enough
to
open
a
shape
file,
should
I
do
and
so
my
work
around
until
I've
solved
the
shape
file
problem
is
to
just
make
geojsons
for
them,
but
it
is
really
really
important
that
that
shapefile
support
be
good.
A
B
B
B
B
I
just
call
it
the
unpronounceable
four-letter
thing,
which
is
great,
and
you
know
it
it
easily
converts
between
all
the
different
kinds
of
things
you
can
either
do
it
on
the
command
line
or,
if
you're,
feeling
lazy,
because
sometimes
it's
hard
to
remember
what
comes
first
in
the
ochre
to
ogr
syntax.
Sometimes
I
just
fire
up
qjs
and
do
it,
but
yeah
it
is
the
shape
file.
B
Is
this
thing
which
everyone
just
never
wants
to
fix
because
it's
such
a
pain
in
the
butt
but
yeah
so
anyway,
so
I
have
got
part
of
the
the
spec
implemented
to
read
the
shp.
That's
pretty
good!
The
reason
why
I
stopped.
I
remember
the
reason
why
I
stopped
the
the
dot
prj
file,
the
projection,
the
w.
When
you
read
that
in
it
the
shape
files
basically
can
define
as
many
unique
kinds
of
projections
as
they
want.
B
They
don't
only
have
to
be
eps
g
codes,
they
can
be
variants
of
it,
and
so
this
is
where
I
this
is
where
I
started
down
that
entire
rabbit
hole
of
trying
to
find
better
c
bindings
to
the
proj
library
is
because
to
to
accurately
and
correctly
read
shape
files
in
the
jvm
needs
to
be
able
to
correctly
reproject
every
shape.
Every
prj
file
into
a
good,
logical,
eps
g,
which
means
that
that
project
and
I
are
like
better
projection
support-
is
a
prerequisite
for
me
solving
shapefiles.
B
A
Right,
I
I
personally
have
a
question,
so
I'm
asked
what
other
than
like
you're
saying
that
documentations
are
your
criminal
night
so
other
than
documentation
like
how?
How
else
can
we
help
you
in
like
building
up
this
gis
ecosystem,
like
you're
saying
like
we
can
also
do
what
requests,
but
I
think
there
are
certain
gaps
here
and
there
that,
like
like,
for
example,
the
clj
part
for
geo,
I'm
not
sure
whether
or
not
we
can
see
or
progress
with
that
currently.
B
Yeah
yeah,
so
what
I
need
to
do
is
I
should
just
like
basically,
okay,
so
I
think
I've
mostly
got
it.
The
reason
why
I
haven't
made
a
pull
request
is
I
keep.
I
keep
doubting
myself
and
I
keep
just
going
in
circles
with
the
internal
structure
and
I
don't
want
to
make
some
messy
pull
requests.
It's
got
like
5
000
commits
that
are
just
going
in
a
circle,
so
that's
my
own.
It's
my
own
shyness
to
actually
pull
the
thing
out
there,
but
basically
here's
what
I
think
I've
done.
B
So
I
now
that
I'm
saying
it
on
a
talk,
that's
about
to
be
on
youtube.
I
hope
that
this
will
cause
me
enough
personal
motivation
to
actually
get
the
pull
request,
kind
of
posted,
but
basically,
what
I
can
do
is
I
I
finalize
it
so
that
basically,
the
the
weird
interoperability
that's
unique
to
jts
and
jsts
have
to
live
in
weird
new
name
spaces
and
I've
kind
of
I
think
I've
mostly
reorganized
that
if
it's
okay,
then
what
whoever
would
want
to
help
is
going
to
have
to
stare
at
some
incomplete
pull
requests.
B
B
The
main
problem
for
doing
the
port,
also,
which
has
been
a
real
pain,
has
been
that
the
tests
are
all
in
midge
right
now,
which
is
a
testing
framework
in
closure
which
does
not
work
in
closure
script,
so
the
port
actually
requires
translating
the
entire
test
suite
into
closure.test,
and
the
only
thing
that's
worse
than
writing.
Documentation
is
rewriting
tests
from
one
testing
framework
to
another.
So
I
apologize
that
you
know.
That's
just
like
this
thing
that
every
time
I
start
to
do
it,
I
always
find
something
better
to
want
to
do.
B
It's
definitely
been
a
factor
in
my
not
completing
this
pull
request,
but
I'll
at
least
open
an
incomplete
one
where
the
tests
don't
fully
pass,
and
you
all
can
maybe
at
least
start
looking
at
it
and
tell
me
telling
me
if
it
kind
of
makes
sense,
and
then
you
know
that
at
least
will
be
functionality
that
exists
on
a
branch,
even
if
the
test
suite
isn't
fully
finished
yet
so
I
I
will.
I
will
do
that.
C
I
had
a
question
about
like
the
geos
like
jts,
like
connection
I
was
actually
I
was
actually
somewhat
surprised
to
to.
I
mean
I
think
I
I
probably
previously
know
that
that
gs
was
a
c
c
port
of
of
jts,
but
you
know
that
kind
of
I
guess
like
based
on
my
cursory
googling.
It
doesn't
look
like
you
know
like
the
port
of
of
gts,
that's
in
geo,
so
it
takes.
C
Right,
like
I
think,
like
it's
like
chris
christopher's
work
like
you,
can
you
can
hook
into
like
the
c
abi
of
something?
And
if
there's
like
a
c
library
out
there
that
has
like
vectorized
operations,
you
could,
even
you
could
even
you
know,
beat
the
performance
of
gts.
You
know
from
closure
by
hooking
directly
into
that
into
that
into
that
c
library.
B
I
don't
know
of
any
alternatives.
I
think
part
of
it
is
also
just
that
geos
history
predates
all
that
kind
of
stuff
and
it's
used
in
so
many
places
that,
and
I
I
want
to
be
really
clear-
I
am
absolutely
not
a
c
programmer,
so
I'm
a
little
bit
static
outside
of
my
wheelhouse,
but
I
think
you
know
backwards.
Compatibility
is
a
really
big
thing
and
it
is
such
a
fundamental
piece
of
so
many
libraries
that
my
guess
is
anything
that
implements
these
newer
kinds
of
features
isn't
going
to
be
geos.
B
B
B
So,
actually,
there's
been
a
fair
uptick
in
the
last
couple
years
of
adding
stuff
into
gts,
which
has
always
been
a
pretty
stable,
bedrock
kind
of
library
but
yeah
to
your
point
about
other
kinds
of
things
that
sort
of
use
the
graphics
card
and
these
other
kinds
of
operations
there
are
more
optimized
optimizations
like
that
is
all
brilliant.
That
is
the
wild
west,
and
if
you
want
to
take
a
stab
at
it,
I
think
you'd
make
a
lot
of
people
happy
it's
possible.
B
You
know
you
can
record
you
can
implement
apache
arrow
with
that
as
a
way
to
sort
of
have
large
scale.
Data
sets
stored
and
sort
of
operate,
those
in
more
efficient
kinds
of
ways,
sort
of
parallel
to
what
you're
talking
about
and
there's
a
lot
of
work
right
now
happening
in
python
to
try
to
even
just
think
through.
How
can
we
store
spatial
objects
in
the
sort
of
geos
world
of
coordinates
and
coordinate
sequences
and
geometries
in
apache
arrow?
B
That
kind
of
makes
like
no
one's
even
solved
that
yet
I
do
think
that
if
they
do
that'd
be
great,
because
that
would
line
up
pretty
well
with
jts.
That
would
mean
that
tech
ml,
potentially,
if
you're
reading
it,
maybe
you
could
think
through
it.
But
this
is
not
a
solved
problem,
so
it
certainly
bears
thought.
It's
probably
not.
The
first
thing
that
I'm
going
to
be
thinking
of
since
I'm
still
trying
to
solve
just
like
a
basic
closure
script
port.
But
I
would
encourage
you
to
take
a
look.
B
D
I'm
also
you've
been
using
h3
a
lot.
It's
super
valuable
one
thing.
I'd
love
to
be
able
to
do
is
to
figure
out
some
way
of
for
for
a
particular
h3
hexagon
at
a
particular
resolution.
What
is
the
population
in
that
hexagon?
Any
I
mean
any
advice
you
can
give
me
on
how
I
could
go
try
to
solve
that
problem.
B
B
The
united
states
all
right?
So
let's
say
you
had
your
either
all
the
united
states
or
some
subset
of
it
right.
You
can
actually
get.
You
can
get
a
well
pretty
soon,
assuming
the
federal
government
doesn't
totally
collapse,
we'll
have
the
2020
census,
but
in
the
meantime
we
have
the
2010
census.
The
census,
by
the
way
for
those
who
are
not
americans,
is
in
a
decennial
count
of
every
single
person
in
the
united
states.
B
They,
actually
you
fill
out
a
survey
or
people
are
sent
out
to
every
individual
household
in
america
and
they
ask
you
how
many
people
live
there
and
some
basic
information
right
and
it's
released
publicly.
You
can
get
for
the
2010
census
from
the
census
bureau,
a
shapefile
of
at
the
block
level,
which
is
the
most
accurate
level
of
of
or
most
fine-grained
level
of
census,
information,
the
population,
the
number
of
housing
units
and
some
information
about
households.
B
If
you
want
to
get
better
stuff
about
other
kinds
of
demographics,
educational
attainment,
other
kinds
of
stuff,
you
have
to
go
out
to
the
block
group
or
census
tract
level
and
do
a
whole
bunch
of
sort
of
census
stuff.
But
you
can
get
for
10
years
ago,
which
is
not
perfect,
but
to
start
you
know,
block
level
data
for
every
block
in
america
by
state
you
can
download
a
shapefile
of
that.
You
can
save
it
as
a
geojson
right
via
ogr
to
ogr,
right
now
or
eventually,
maybe
via
a
closure
library.
B
So
you
have
a
contiguous
perfect.
You
know
for
the
state
of
california.
I
think
in
your
case
right
it
can
take
a
perfect.
You
know
contiguous
topologically,
consistent
block
level
population
count.
What
you
can
then
do
is
for
an
h3
cell.
You
do
the
intercepts
right.
So
first,
your
question
would
be
for
an
h3
cell
find
which,
via
if,
when
st
intersects
or
when
dot
intersects
in
a
gts
kind
of
way,
which
blocks
does
it
cover
great,
so
it'll
be
one
or
perhaps
more
than
one,
because
polygons
touching
polygons,
you
might
have
multiple
blocks.
B
At
that
point,
you
need
to
kind
of
guesstimate
how
many,
what
proportion
of
the
block
that
the
h3
cell
is
covering
is
there,
and
so
what
you
would
do
is
let's
say
you
go
to
h3
cell
and
it
covers
it.
Somehow
overlaps
two
different
census
blocks.
You
would
find
the
two
geoids
of
the
two
different
census
blocks
that
have
a
population.
B
You
could
say:
okay,
great
this
enter
you
do
the
intersection
function,
whether
it's
an
st
intersection
or
in
geos
or
avid
or
aurelius,
or
something
you
know,
here's
the
area
that
is
common
to
both
the
h3
cell
and
you
know
that
census
block
and
then
you
would
need
to
do
an
area
comparison.
This
sort
of
partial
h3
cell
on
this
census
block
is
15
or
20
of
the
block
itself.
You
could
say
maybe
0.2
or
0.15
proportion
of
that
block
equals
this
many
people.
That
would
be
the
urban
planar
method
to
guesstimating
it.
B
If
you
want
to
get
really
fancy,
which
I
bet
you
don't
what
you
could
do
is
do
another
level
of
gis
analysis.
You
could
download
the
parcel
information
for.
I
think
you're
in
san
francisco
right,
so
you
could
download
san
francisco
parcel
information
right,
the
actual
property
information
for
the
city
of
san
francisco.
B
You
could
then
do
all
that
crazy
stuff.
I
was
saying
about
the
zones
and
all
the
different
kinds
of
things
you
could
limit.
You
could
find
only
the
residential
land
in
san
francisco
or
mixed
use,
where
sometimes
you
have
like.
Maybe
you've
got
like
your
prep
on
the
first
floor
or
like
a
panerai
in
the
first
floor
and
then
the
apartments
above
right.
You
can
find
all
that
stuff
out,
so
you
can
actually
get
down
to
the
property
level.
If
you
really
really
wanted
to
get
very.
B
You
could
say
find
only
land
where
residential
stuff
can
be
on
each
census
block.
So
then,
instead
of
just
having
a
contiguous
census
block
covering
water,
open
space
roads,
everything
else
with
the
number
of
people,
you
could
even
confine
it
to
make
it
more
accurate,
only
residential
land
on
every
census
block.
B
And
then,
if
you
have
your
partial
cell,
you
can
be
more
confident
that
your
little
proportion
kind
of
represents
only
residential
land,
for
example,
if
that
h3
cell
overlapped
with
you
know,
you
know,
just
a
park
in
san
francisco
like
there'd,
be
no
people,
and
so
that
would
be
a
way
to
be
more
accurate
if
you
wanted,
but
all
these
things
that
I'm
saying
just
for
everyone
else's
reference.
These
are
not
fancy
cardo
level
kinds
of
tools.
B
D
D
E
B
B
All
right
so
the
first
is
it
was
the
don't
reinvent
the
wheel.
I
don't
want
to
write
qjs
again
like
that
tool
is
amazing.
If
it's
just
basic
map
making
don't
reinvent
it
it's
the
best,
it
is
better
than
anything
we're
gonna
write.
Thousands
of
people
are
contributing
to
on
a
regular
basis
for
static
maps.
B
Similarly,
at
the
moment,
mapbox
gl
and
by
extension,
sort
of
uber
things
over
it
right
that
dot
so
sorry,
maps
gl
is
a
framework.
Is
a
javascript
library
that
that
uses
webgl
to
make
slippy
maps.
So
it's
not
google
maps,
it's
not
leaflet
it's
this
other
one
and,
to
my
mind,
I
at
the
moment
have
not
found
a
better
thing.
I
would
call
it
state
of
the
art
for
performance
web-based
mapping.
B
It
does
a
lot
of
things
really
really
well.
You
can
do
shockingly
advanced,
styling
and
visualization
of
base
maps.
You
can
color
every
single
layer
with
all
kinds
of
different,
complicated
things,
there's
even
tools
called
like
mapbox
studio
and
there's
like
offline
versions
and
there's
open
source
alternatives
to
create
these,
like
basically
json
style
jsons.
B
That
can
modify
every
aspect
of
the
base
map.
Additionally,
you
can
add
in
your
own
geojson,
or
you
can
add
in
your
own
vector
tiles
that
you
can
generate
in
postgis,
or
you
know
kind
of
in
java
that
you
can
then
add
on
to
it.
So
you
can,
via
macbox
gl,
put
a
lot
of
really
fancy
stuff
into
interactive
web
javascript
mapping.
On
top
of
that,
there's
a
set
of
other
tools
that
you
know.
I
think
some
of
you
on
this
call
are
using
as
well
there's
one
called
deck.gl
which
is
written
by
uber.
B
It's
not
h3,
but
I
think
it
does
some
hexagonal
analysis,
but
it
also
it's
not
just
sort
of
flat
stuff
on
matbox
gl,
it's
a
whole
set
of
sort
of
additional
visualization
tools
that
live
on
top
of
mapbox.
So
between
that
between
mapbox,
I
think
the
javascript
tools
are
like
this
sort
of
other
pet,
like
qjs,
is
the
best
for
just
static
maps.
If
you
want
to
make
a
pretty
advanced
map
across
all
kinds
of
different
projections,
that's
your
thing.
B
If
you
want
to
make
web-based
visualizations
that
are
more
dynamic,
I
can
do
things
in
three
dimensions
that
can
do
things
that
are
much
more
sort
of
statistically
interesting,
mapbox,
gl
and
decgl
are,
are
the
best
and
and
there's
no
reason
to
reinvent
them,
so
to
the
degree
that
you
can
as
a
closure
scripts
developer
just
integrate
with
those
two
for
the
web.
Like
you're
good,
like
I
think,
that's
solved
right
or
that
piece
of
it
is
solved.
The
question
becomes
what
else
there
is
to
do.
B
That's
what
initially
made
me
start
this
this
never
ending
test
of
the
closurescript
port,
which
is,
I
feel
like
for
me
at
least
right
now.
You
know
I
do
a
bunch
of
coding
in
my
repple
I
make
a
bunch
of
random
shape
files
or
geojsons
or
whatever
they
are,
and
I
do
analyses
I
do
my
transducers.
I
make
a
new
thing.
I
export
it
out.
B
I
put
it
in
put
in
maybe
I
upload
to
postjs,
maybe
I
put
it
in
qjs
and
make
a
map
in
a
slightly
alternative
world
where
I
were
still
making
web
maps
all
the
time.
I
could
be
doing
it
in
mapbox
right,
but
there's
still
at
the
moment
a
lot
of
like
closure
to
uploading
it
to
mapbox,
to
whatever
else
I
think,
the
claudia
script
port,
I
think,
would
would
be
helpful
because
it
would
mean
that
you
can
do
some
of
that
dynamic,
visualization
and
analysis
up
in
the
front
end
as
well.
B
I
think
that's
useful
now,
I'm
not
a
full-time
floaterscript
developer.
It's
also
part
of
my
problem
is
like
it's
like
I'm
having
to
learn
all
the
the
nuances
of
javascript
and
clojurescript
that
I'm
not
using
all
the
time.
That's
one
step,
but
I'm
not
sure
what
else
would
be
useful
is
it
is
it
another
kind
of
like
maybe
it's
some
weird
intermediate
between
qgis
and
web
mapping,
something
that
allows
you
to
play
with
some
of
the
tools.
B
I
don't
know,
that's
a
good
question
for
you
all,
but
at
the
moment
at
least,
I
think
sort
of
if
you
find
a
way
to
get
to
mapbox
use
an
app
box.
If
you
find
a
way
to
get
to
qjs
use
qgis,
I
I
it
seems
to
me
clear
that
the
cleric
report
is
useful,
but
for
me
at
least
the
problems
of
visualization,
I
my
needs
at
the
moment
are
solved,
and
so
I'm
curious.
Where
else
you
all
are
finding
pain
that
you
can't
solve
with
existing
things.
D
D
A
web
application
called
kepler.gl,
which
is
a
web
app
that
that
runs
deck
gl
and
it's
sort
of
designed
for
people
who
are
data
scientists
and
have
they
just
have
a
data
set
and
you
can
load
the
data
set
into
kepler
and
hence
deck
gl
and
then
start
playing
around
with
layers
and
and
and
visualize
that
way.
So
that's
sort
of
the
no
programming
model
of
visualizing
some
data.
Now
you
know
I'm
making
a
joke
about
the
prayer.
D
Sometimes
you
have
to
program
to
get
the
data
in
the
right
format
to
load
it
into
kelper.jl.
So
we
did
that.
That's
where
I
started,
and
then
we
did
that
it
worked
great
and
I'm
like
okay,
this
is
cute,
but
I
don't
like
to
like
it
didn't
know
anything
about
my
domain.
You
know
kelp
was
just
like
you
know:
you've
got
you
know
json
and
data
and
layers,
and
this,
and
so
I
wanted
to
have
something
more
domain
specific.
D
So
then
I
jumped
all
the
way
over
to
deck
gl
and
created
a
web
application
and
in
the
combination
of
closure
enclosure,
script
that
was
specific
to
my
domain
type
in
the
database.
So
that
was
very,
very
powerful.
Now
the
the
next
thing.
So
then,
at
some
point
in
that
whole
process,
the
the
uber
people
came
out
with
a
a
python
binding
to
deck
which
in
the
pi
deck.
D
So
then
you
can
it's
sort
of
an
intermediate,
it's
sort
of
an
image,
so
you
can
sit
there
in
python
and
then
define
deck
layers
and
and
tied
in
with
jupiter.
I
think
all
that
stuff.
So
that
was
a
cute
idea
and
but
I
really
want
to
do
that
from
closure,
so
I've
been
working
with
a
colleague
on
sort
of
the
deck,
so
here's
so
the
deck
gl
binding
in
in
the
browser
from
the
closure
rebel.
D
So
the
the
thing
that
you
all
probably
know
about
is
eyes,
which
is
basically
that
for
vega,
so
for
vegans,
so
in
the
closer
ruppel
you
could
pull
up
a
browser
window
and
start
doing
vega
kinds
of
things
from
closure.
So
I'm
trying
to
to
do
that
for
dec
gl
and
we
have
early
versions
of
that
are
starting
to
work.
So
that's
pretty
powerful
too.
So
I
can
throw
together
a
geospatial
visualization
from
the
ralph
hall
and
play
around
with
it.
F
That
sounds
really
cool,
and
I
would
be
interested
to
hear
much
more
about
that
as
well.
What
I
was
trying
to
say
earlier
was
that
that
there
is
a
map
box
integration,
also
in
the
in
the
microsoft
power
bi
data
analysis
tool
that
lets
you
easily
mix
and
match
different
kinds
of
data
sets
and
and
draw
a
map
out
of
them
and
visualize
them
it.
I
tried
it
in
the
in
the
summer
and
it
wasn't
quite
sophisticated
yet,
but
but
in
the
github
it
looked
like
that.
F
It's
it's
under
development
and
that
might
turn
out
to
be
something
something
really
useful
for
for
data
scientists
as
well,
and
it
was
quite
simple
to
use
as
well
and
personally,
I'm
very
interested
in
these
exploratory
tools,
where
you
can
just
put
your
data
in
and
try
to
visualize
it
in
different
ways
and
try
to
understand
the
data
better.
So
that's
why
I
was
very
enthusiastic
about
about
done
your
what
you
said.
That
so
is
that
something
something
open
source
you're
working
on
or
or
is
it.
D
A
People
probably
want
to
hear
but
but
yeah
so
so
for
now
like
for
those
that
need
to
go
off
we'll
sort
of
be
ending
the
formal
session
here.
But
we
can't.
We
can
definitely
continue
the
discussion
here
and
yeah,
but
yes,
so
huge
thanks
to
will
for
all
this
time
his
time
and
effort
to
talk
about
clojure's,
geospatial
space
and
all
his
insightful
thoughts
and
comments.
Yeah.
F
Yeah,
first
of
all,
thanks
for
the
great
presentation
will
I
can
all
already
almost
see
in
my
eyes
that
I'm
I'm
evaluating
those
those
cool
like
I'm
having
my
geospatial
data
and
I'm
flowing
through
the
dance
transducers
and
then
rendering
them
to
a
web
map
in
enclosure
script.
So
that's
that's
like
something
that
I
would
like
to
see
in
the
future
happen
in
in
in
a
very,
very
rebel-driven
way.
B
B
Okay,
I'll
get
it
how
about
this?
I
will
get
it
to
a
point
where,
where
I
am
right
now
is
a
good
starting
point
and
then
like,
I
think
the
main
thing
I'm
going
to
need
help
on
is
just
some
of
those
tests
and
just
make
sure
I'm
dotting
all
my
eyes
and
crossing
my
t's,
so
I'll
put
it
there
and
I'll
start
it
and
then
it'll
be
on
the
factual
geo
repository
and
so
we'll
figure
out
a
way
to
sort
of
collaborate
and
get
that
thing
done.
D
One
of
you
know
I
I
didn't
I
was
previously.
I
didn't
quite
understand
where
you're
coming
from
with
that
feature
like,
but
now
your
explanation
was
excellent,
so
now
I
get
it
and
now
I'm
excited
about
it.
I
think
that
for
me
you
know,
I
have
a
different,
I'm,
not
an
urban
planner,
I'm
dealing
with
aircraft
and
aircraft
trajectories.
D
D
B
Well,
so
the
newer
versions
of
jts
in
the
factual
library,
like
one
of
the
changes,
is
it's
not
perfect,
and
this
is
where
this
is
where
actually
backwards.
Compatibility
became
a
bit
of
a
complication.
I
went
ahead
and
you
know
when
we
added
the
projections
and
stuff
into
the
actual
geo.
It
was
like.
B
So
the
the
sort
of
the
native
geometry
generation
features
of
the
factual
geo
library
do
not
get
the
m
very
well.
However,
jts
now
has
a
fourth
dimension
in
coordinates,
available,
they're
they're,
only
kind
of
exposed,
but
they're
called
coordinate
like
xyzms,
basically,
and
you
can
construct
geometries
out
of
them
and
you
can
and
so
like
they
live
inside
all
of
the
different
jts
objects.
B
So
actually,
what
might
be
useful
if
you're,
if
you'd
be
willing,
is
if
you
could
just
poke
around
a
little
bit
in
like
try
it
out
a
little
bit
in
the
jts
and
figure
out?
Is
there
any
mistake
like?
Did
we
screw
anything
up
in
factual
geo
where,
like
it's
incorrectly
dropping
ends
when
it
shouldn't?
I
think
at
the
moment,
if
you
put
ends
in
correctly
they're
there,
it
doesn't
lose
any.
B
But
what
you
know
using
avid
and
aurelius
might
be
a
good
chance
to
kind
of
revisit
some
of
that
api
to
to
get
it
a
little
bit
more
consistent
like
it
can
still
live
and
keep
backwards,
compatibility
with
ngo
just
fine
like
I
don't
want
to
break
anything,
but
we
that
that'd
be
actually
that'd,
be
something
which
I'd
like
to
think
through
a
little
bit
more
consistently
before.
I
do
a
first
alpha
release
because
I'd
like
avid
and
aurelius
to
better
think
about
it.
B
D
I'll
I'll
do
that
sure
for
sure
the
other
thing
I'll
just
give
you
a
pen
on
the
back
too,
is
that
I
I
did
poke
in
your
library
that
readable
column
thing
for
the
postgis,
and
I
actually
did
that
because
I
I
do
have.
I
have
some.
You
know
trajectories
in
postgis
that
have
the
time
because
you
know-
and
I
was
able
to
read
it
in
to
closure,
so
that
was
I
was
pretty
excited
about
that.
D
B
No,
it's
a
totally
good
question.
A
coordinate
is
a
projectionless
thing.
It's
just
a
number
of
it's
an
x,
a
y,
a
z
and
potentially
an
m
that
just
exists
in
in
no
plane
whatsoever.
A
point
is
a
a
single
coordinate
projected
somehow
or
projected
with
no
projection.
It
could
have
a
projection
of
zero
right.
A
a
line
string
is
a
geometry
with
multiple
points
you
know
with
sorry
with
multiple
coordinates,
so
this
is.
C
B
Where
you
know
a
poly
like
a
point,
has
a
single
coordinate
sequence
of
one
point:
a
line
string
has
a
single
coordinate
sequence
of
like
a
vector
of
of
of
coordinates.
A
polygon
has
potentially
multiple
line
strings
which
are
themselves
individually,
coordinate
sequences.
A
multi-polygon
is
that
in
multiple
different
polygons
put
together,
but
deep
down
inside
they're
all
coordinates,
and
so
the
whole
way
that
proj
4j
or
the
geo
and
project4j
work
together
is
basically
when
you
reproject
a
point.
B
It's
more
like
that's
the
little
internal
thing
within
a
geometry
like
you,
modify
a
coordinate
and
reapply
it
to
a
geometry,
that's
kind
of
how
projections
work,
so
the
main
flaw
in
terms
of
the
the
the
m
or
t
coordinate
might
be
if
there's
a
mistake
somewhere
and
that
praj4j
doesn't
do
a
very
good
job
of
thinking
about
t.
So
I
can't
remember
if
praj4j
drops
the
t
or
not,
but
I
do
know
that
sort
of
factual
geo
on
its
own
and
gts.
B
If
the
t's
in
there,
it
should
stay
it's
only
if
you
do
something
where
it
might
be
messing
up
the
coordinates
and
dropping
the
t.
That's
and
actually,
if
you
to
answer
the
question
better,
I'm
not
going
to
do
it
right
now,
but
if
you
look
in
the
factual
geo
test
suite,
I
did
address
this
in
some
level
like
there's
like
a
note
where,
like
the
test
is
like
you
know,
when
you
reproject,
it
drops
the
t,
but
when
you
make
shapes-
and
you
turn
it
into
a
thing-
it
keeps
it
like.
B
D
Yeah,
so
that
I'm
super
motivated
so
I'll
definitely
take
a
look
at
that
and
get
in
touch
with
you
just
not
that
it
matters
that
much.
But
you
know
you're,
like
you
care
a
lot
about
intersex
and
all
this
kind
of
stuff.
I
get
it
from
your
work.
So
what
the
thing
that
I
really
care
about
with
these
trajectories
is:
what's
the
point
and
time
of
cl
and
distance
of
closest
approach
between
a
location
of
interest
and
a
a
trajectory
so
like?
When
did
this
airplane
come
closest
to
my
house?
D
Jts
has
those
so
yeah
anyway.
I
I
figured
out
with
with
so
long
story
short.
I
basically
post
just
has
a
bunch
of
stuff,
that's
helpful
for
this.
They
have
a.
They
have
3d
closest
point
and
things
like
that
and
and
works
with
trajectory.
So
I
I
had
to
go,
seek
help
from
post
just
experts
so,
but
I
I
was
able
to
get
seek,
got
that
help
paid
for
the
privilege,
but
it
was
reasonable
and
then
now
I
can
actually
get
out
of
a
post.
Just
query
that
it's
the
worst
query.
D
I've
ever
seen
that
they
wrote
for
me,
but
it
can
return
the
point
of
closest
approach,
the
time
clock,
sports
and
business
close
work,
which
is
super
powerful.
He's
like
give
me
all
the
aircraft
that,
within
that
at
any
point,
a
trajectory
came
within
three
kilometers
of
my
house
and
what
time
did
each
one
of
those
ones?
Aircraft
come
closest
and
you
know
did
how
far
away.
B
So
how
about
this?
What
your
job
then
would
be
and
I'd
be
happy
to
have
trouble.
Well,
I
I'm
committing
to
too
many
things.
I
also
have
a
way
of
over
promising,
but
what
you
know,
what
might
be
good
is
to
figure
out
where
jts
starts
to
diverge,
because
they
are
trying
to
increase
that
functionality,
and
so
I
know
that
jts,
for
example,
has
3d
coordinates
and
4d
coordinates,
so
those
things
should
be
able
to
exist,
and
you
know
if
closest
point,
you
know
if
a
lot
of
the
2d
ones
exist
in
post
js.
B
My
guess
is
some
version
of
that
exists
somewhere
deep
in
the
vowels
of
the
algorithms
in
jts
for
2d,
it's
possible
3d,
isn't
there.
So
the
right
step,
you
know,
is
just
figuring
out
where
it
fall
like
where
it
fell
in
between
post.js
and
whatever.
It
may
just
be,
adding
a
couple
out
like
fairly
small
algorithms,
whatever
the
functionality
is
in
geos
to
be
back
in
jts,
and
then
that
should
work
all
the
way
back
up
the
chain,
we
should
have
it
in
the
jvm.
D
That
sounds
great
yeah.
That
sounds
great
I'll,
do
a
bunch
of
work
and
then
I'll
contact
you.
You
can
give
me
some
advice
and
we
can.
You
can
point
me
in
the
right
direction
and
I'm
happy
to
do
some
work
on
this
to
support
it
because
I
care
about
it,
but
it
would
be
great
to
have
somebody
who
you
know.
You
have
a
lot
of
experience
with
the
geo
library
which
I
don't
have
and
the
internals
and
stuff
like
that.
D
A
A
B
I
would
say:
oh
really
needs
more
avid's,
pretty
small,
it's
actually
just
like
it's
really
just
the
feature
like
I
want
to
make
one
library,
that's
just
a
plain
abstraction.
Maybe
you
don't
need
all
the
fancy
stuff
about.
You
know
like
all
the
different
transducers
and
things,
so
I
just
wanted
to
create
very
one,
very
small
library,
because
people
might
want
to
be
able
to
use
feature
like
without
using
all
the
other
stuff.
So
that's
why
I
made
it.
It's
pretty
small.
I
don't
think
there's
that
much.
B
That
needs
to
go
there
as
far
as
I'm
concerned.
If
you
want
to
start
poking
around
aurelius,
basically,
here's
where
documentation
documentation
could
go.
You
look
at
like.
I
have
no
idea.
Why
will
wrote
all
this
like
even
just
saying
I
don't
know
why,
like
it's
it's
self-evident
to
me,
because
I
wrote
it
so
it
would
be
helpful
just
to
be
like
this
part
doesn't
make
sense.
I
don't
understand
the
use
case.
Here's
some
things
I
could
try
out.
B
B
I
use
it
all
the
time,
but
I'm
trying
to
make
it
easier
for
planners
and
or
not
planners,
but
for
the
analysis
too,
and
so
I
would
love
for
you
know
it
to
get
a
little
bit
more
stabilized,
get
to
a
first
release
and
then
it
get
built
upon
as
a
good
exploratory
tool.
So
I
guess
you
should
come
talk
to
me
and
we'll
make
a
plan.
A
If
anyone
has
other
questions,
please
interrupt
me,
but
another
question
that
I'm
personally
sort
of
curious
is
that
I'm
familiar
with
the
sort
slightly
familiar
with
like
rhino
and
like
grasshopper
space.
But
I
was
thinking.
Isn't
there
like
a
similar
idea
with
like
arcgis,
so
arcgis
also
has
like
sort
of
like
visual
scripting
yeah.
So
does
that
also
sort
of
fulfill
the
criteria
you're?
Looking
at
for
grasshopper.
B
Sort
of
they
have
that
in
qjs2
it's
it's
called
like.
I
don't
even
use
that
it's
called
like
modeling.
There's
like
these,
these
things
where
you
can
take
the
different
operator,
it's
actually
very
similar
to
what
I'm
talking
about
buffer
and
whatever
and
whatever
and
you
chain
them
together,
and
you
can
make
these
little
formulas,
but
there's
major
downsides.
B
Well,
they
work,
so
they
work.
But
there's
a
couple
major
downsides:
qjs
in
arcgis
are
not
as
performant
as
pose.js.
They
just
you
know
single
threaded
operations
that
go
one
after
another,
there's
a
lot
of
very
understandable
baggage
that
comes
with
using
these
tools
to
get
the
stuff
done.
It
gets
filtered
off.
You
know
it's
the
part
of
the
power
I
have
found
in
post
gis
and
also
with
recent
releases
of
postgresql.
B
So
you
know
that's
a
much
harder
problem
to
solve
in
qgis
than
it
might
be
where
we
think
it
through
with
that,
you
know
as
a
as
a
as
a
basic
principle,
and
the
thing
is
transducers
and
a
lot
of
things
about
closure
set
the
stage
for
that.
So
I
feel
like
it's
a
good
base
to
build
it
on,
but
you're
absolutely
right.
Those
things
do
exist
in
qjs
already
I
just
they
don't
solve.
B
They
make
my
own
for
some
reason
something
about
the
use
of
them
feels
harder
than
it
should
be,
but
the
functionality
is
abs
is
absolutely
there
in
some
form.
I
I
it's
very,
very
powerful.
I
just
it
something
about
it
feels
like
it
makes
my
life
harder,
so
I
don't
find
myself
drawn
toward
it
for
some
reason
or
another.
A
E
Maybe
I
wish
to
suggest
something:
please
tell
me
if
it
sounds
good,
so
it
could
be
nice
to
have
some
visualization
way
as
a
tool
for
documentation
so
that
we
can
have
a
certain
walk
through
with
some
operations
and
see
them
in
the
docks,
and
for
this
we
may
choose
possibly
not
the
best
way
of
visualization,
but
one
of
the
easiest
way,
maybe
leaflet
or
something
like
that
and
and
then
just
try
to
to
write,
maybe
together.
If
you
wish,
we
can
meet
and
write
stories
and
see
if
we
can
find
out
some
rebel
experience.
E
E
B
I
have
a
suspicion,
it's
the
kind
of
thing
that
sounds
small,
but
gets
really
really
big.
Like
you
know,
I
I've
started
to
think
about
this.
The
last
couple
months
right,
I've
been
I've,
been
teaching
this
course
for
a
couple
years
now
to
undergraduates,
and
I
have
a
lot
of
advice.
So
I've
got
my
own
personal
workflow
for
how
to
make
this
stuff
make
sense.
B
I've
learned
a
lot
over
the
last
couple
years
of
how
people
who
are
new
to
it,
where
their
pain
points
are,
but
that
still
isn't
even
a
standalone
set
of
documentation
and,
as
you
know,
my
this
higher
education
is
having
to
figure
out
a
way
to
be
more
distance,
I'm
having
to
modify
the
way.
I
think,
through
these
kinds
of
things,
at
the
very
b,
a
even
further
dream
than
like
finishing
some
of
these
small
library
tools
is.
I
think
I
would
like
to
try.
B
Trying
to
expl
trying
to
teach
this
whole
field
is
such
a
daunting
thing.
It
really
is
like
a
like
a
textbook
right
like
in
the
end
for
to
that
end,
where's
my
copy,
if
nothing
else,
there's
a
textbook
called
post
gis
in
action.
It's
it's.
It
is
a
textbook
written
by
the
writers
and
post
gis.
If
you're
actually
just
curious
in
learning
more
it's
a
good
book
like
you
should
get
that
textbook.
It's
not
closure.
It's
not
you
know,
but
it
is
an
exhaustive
and
comprehensive
look
and
how
to
do
complicated,
valuable.
B
You
know
well
written
like
sql
queries
and
post
jf.
It's
a
good
place
to
start
and
in
a
weird
way,
it
kind
of
can
be
some
advice
for
what
a
different
version
that
might
be
closure,
oriented,
might
kind
of
want
to
be,
but
for
people
who
want
to
know
more
right
now,
I
would
encourage
you
just
to
look
at
that
textbook.
It's
the
best
resource.
I
know
right
now
about
gis
analysis,
especially
in
a
programmatic
way.
F
And
I'm
quite
familiar
with
both
gis
and
and
and
q,
q,
gis
and
these
tools
and
so
on.
But
but
I'm
thinking
something
like
having
having
a
subset
of
those
features
running
running
in
a
browser
so
that
a
non-gis
expert
could
make
sense
of
of
their
spatial
data.
And
that's
that's
something
that
that
I
would
like
to
work
towards.
F
And
I
I
really
like
that,
like
that,
your
idea
of
having
those
having
those
simple
spatial
operations
like
intersect
sections
and
joints
and
that
that
kind
of
things
so
that
if,
if
we,
if,
if
you
can
simply
chain
them
like
this,
then
you
can
easily
describe
that
as
a
data
in
a
declarative
way.
And
then
then
it's
it's
rather
simple
to
have
the
track
and
drop
ui.
Where
you,
where
you
define
those
transformations
you
can.
B
It
first
yeah
like
I
I
I
I
just
I
want
to
just
give
it.
I
want
to
be
realistic.
Like
I
thought
a
year
ago
I
was
like
oh,
this
is
so
easy.
I'm
going
to
put
some
things
together,
I'll
make
some
great
java
fx
thing
well
I'll,
put
all
my
little
transducers
into
little
boxes,
I'm
not
a
great
programmer,
but
like
I
can
make
it
and
then
I
was
like.
Oh
then,
there's
and
then
there's
this,
and
so
that's
why
I
I
framed
it
as
like.
B
There
are
a
lot
of
building
blocks
that
we
have
the
opportunity
to
work
on
making
first,
if
they
all
get
in
place,
the
right
way,
whether
it's
in
javascript
or
java
or
whatever,
maybe
it
could
exist
and
then,
if
that
thing
were
to
be
made
which
exist
in
no
language,
that
would
be
a
very
compelling
use
case
for
closure
in
in
this
world.
But
I
I
just
it
is
very
true
that
there's
a
lot
missing
first
and
like
it's
like
the
breath,
it's
like
my
brussels
sprouts
for
documentation,
there's
a
lot
of
things.
B
It's
never
going
to
be
a
fully
performant
multi-um
projection
based
way
of
visualizing
things,
so
there's
a
couple
different
pieces
that
are
absolutely
missing
right
now,
one
some
version
of
that
prod
c
bindings
version
for
javascript
we
gotta
someone
has
to
make
that
like
what
that
means
is
someone's
gonna
need
to.
I
I'm
not
a
very
good
javascript
programmer
use,
mscriptin
to
compile
the
c
library
for
projections
into
some
crazy
javascript
thing
and
then
write
a
javascript
library
which
talks
to
the
sqlite
gets
the
current.
B
Whatever
string
runs
it
through
the
crazy
unscripted
thing
and
then
reprojects
your
points
I
mean,
if
you
did,
that
you
would
have
an
amazing
javascript
tool
that
can
reproduce
at
any
kind
of
point
and
you
get
to
utilize.
You
get
to
stand
on
the
back
of
all
the
existing
amazing
projection
work
in
proj,
but
you
have
to
do
that
right.
Another
thing
you
know:
mapbox
only
does
the
web
mercator
projection
a
good
gis
tool
actually
needs
to
be
able
to
do
all
these
different
ones.
B
B
But
what
might
even
be
good
is
like
just
brainstorming
all
the
hard
things
we
have
to
do
to
make
it
happen,
and
then
just
whatever
skill
sets
everyone
has,
we
begin
tackling
the
chain
of
them
together.
That
could
be
the
path
forward
and
I'm
not
a
good
front-end
developer.
So
I
would
love
to
rely
on
some
of
your
skill
sets
to
make
that
happen.
F
Yeah,
like
I,
I
really
like
the
way
how
you
think
that
you
want
to
have
the
profound
building
blocks
and
on
top
of
those,
then
you
can.
You
can
do
or
solve
any
gis
problem,
but
maybe
the
other
way
to
start
thinking
about
it
would
be
just
first
support
one
one
projection
and
and
then
start
expanding
from
there,
but
and
and
get
something
something
easily
done
in
the
first
place.
E
B
I
agree
with
that
totally
yeah,
so
all
of
your
front-end
developers
I'm
going
to
defer
a
lot
of
sort
of
the
decisions
about
that
kind
of
stuff
to
you,
because
I
don't
know
it
as
well
as
you
I.
I
really
do
live
a
little
bit
more
in
the
analytical
side
of
things,
but
I
I
want
to
help,
however,
is
useful
to
make
that
happen.
F
Yeah,
I
see
it
as
spectrum
like
there
is
the
like
the
very
low
level
and
then
there's
the
analytical
level
and
then
there's
the
presentation
level
and
I've
been
mostly
working
with
the
presentation
and
then
and
how
to
ingest
the
data
or
how
to
get
normal
people
to
understand
the
the
basics
of
gis
and
putting
the
data
and
then
so
on.
I'm
also
interested
in
the
analytical
section,
because
that's
where,
where
the
actual
interesting
need
of
the
of
the
whole
gis
business
actually
is,
is
in.
B
There
is,
I
will
make
one
pitch
for
why
I'm
so
focused
on
trying
to
make
a
cross-platform
version
of
all
these
building
blocks
and
why
I
think
closure
is
useful
there.
Here's
my
pitch,
which
is
qjs,
is
amazing.
It
works
in
on
the
computer
like
if
you
have
a
chromebook
and
you've
got
your
crazy
linux
subsystem
for
whatever
you
can
get
it
to
work
on.
B
Whatever
else-
and
I
do
see,
value
and
qjs
is
never
going
to
do
that,
because
python
doesn't
run
on
these
things,
and
I
do
it
again.
This
is
like
another
one
of
those
dreams
in
the
future,
but
there
is
a
piece
in
me
that
sees
the
ability
to
sort
of
have
closure
work
on
all
the
current
things.
Like
you
said,
you
know
it's
a
spectrum,
some
pieces
of
it
on
the
jvm
other
pieces
for
visualization.
B
That
would
mean
that
our
entire
thought
process,
eventually,
if
your
visualizations
in
javascript,
if
some
of
the
analysis
can
also
be
in
javascript,
if
the
projections
can
also
be
a
little
bit
in
javascript,
then
actually
the
mobile
device
can
be
a
thing
for
gis
too,
and
not
just
a
visualization
tool.
That
is
not
important.
I
don't
do
that
professionally.
I
I
I
know
that
I'm
like
I'm
getting
way
too
up
in
the
clouds
here,
but
just
for
whatever
it's
worth.
B
It's
the
reason
why
I
would
encourage
our
thought
processes
to
try
whenever
possible,
to
keep
closures
to
bring
closure
script
along
just
because
it
would
actually
do
something
very,
very
new
for
the
field
that
that
would
mean
javascript
development
can
and
therefore
mobile
devices
can
do
what
is
right
now
really
stuck
on
desktop
computers,
but
again,
I'm
now
getting
way
up
in
the
clouds.
I
know,
but
it's
just
it's
worth
thinking
about.
F
Yeah
but
open
layers
works
works
really
great
on
on
mobile
mobile
devices,
and
it
has
it
has.
It
has
a
huge
amount
of
functionality
and
it's
under
heavy
still
under
heavy
development.
So
I
I
see
a
lot
of
potential
in
that
and
I'm
using
open
layers
myself
quite
a
lot
for
for
implementing
the
presentation
layer
and
then
doing
simple
gis
operations
in
the
in
the
browser
also
on
the
mobile.