<s>
In	O
computer	B-General_Concept
science	I-General_Concept
,	O
a	O
Judy	B-Data_Structure
array	I-Data_Structure
is	O
a	O
data	B-General_Concept
structure	I-General_Concept
implementing	O
a	O
type	O
of	O
associative	B-Application
array	I-Application
with	O
high	O
performance	O
and	O
low	O
memory	O
usage	O
.	O
</s>
<s>
Unlike	O
most	O
other	O
key-value	B-Data_Structure
stores	I-Data_Structure
,	O
Judy	B-Data_Structure
arrays	I-Data_Structure
use	O
no	O
hashing	O
,	O
leverage	O
compression	O
on	O
their	O
keys	O
(	O
which	O
may	O
be	O
integers	O
or	O
strings	O
)	O
,	O
and	O
can	O
efficiently	O
represent	O
sparse	O
data	O
;	O
that	O
is	O
,	O
they	O
may	O
have	O
large	O
ranges	O
of	O
unassigned	O
indices	O
without	O
greatly	O
increasing	O
memory	O
usage	O
or	O
processing	O
time	O
.	O
</s>
<s>
Roughly	O
speaking	O
,	O
Judy	B-Data_Structure
arrays	I-Data_Structure
are	O
highly	O
optimized	O
256-ary	O
radix	B-Data_Structure
trees	I-Data_Structure
.	O
</s>
<s>
Judy	B-Data_Structure
trees	I-Data_Structure
are	O
usually	O
faster	O
than	O
AVL	O
trees	O
,	O
B-trees	B-Architecture
,	O
hash	B-Algorithm
tables	I-Algorithm
and	O
skip	O
lists	O
because	O
they	O
are	O
highly	O
optimized	O
to	O
maximize	O
usage	O
of	O
the	O
CPU	B-General_Concept
cache	I-General_Concept
.	O
</s>
<s>
The	O
Judy	B-Data_Structure
array	I-Data_Structure
was	O
invented	O
by	O
Douglas	O
Baskins	O
and	O
named	O
after	O
his	O
sister	O
.	O
</s>
<s>
Judy	B-Data_Structure
arrays	I-Data_Structure
are	O
dynamic	B-Data_Structure
and	O
can	O
grow	O
or	O
shrink	O
as	O
elements	O
are	O
added	O
to	O
,	O
or	O
removed	O
from	O
,	O
the	O
array	O
.	O
</s>
<s>
The	O
memory	O
used	O
by	O
Judy	B-Data_Structure
arrays	I-Data_Structure
is	O
nearly	O
proportional	O
to	O
the	O
number	O
of	O
elements	O
in	O
the	O
Judy	B-Data_Structure
array	I-Data_Structure
.	O
</s>
<s>
Judy	B-Data_Structure
arrays	I-Data_Structure
are	O
designed	O
to	O
minimize	O
the	O
number	O
of	O
expensive	O
cache-line	O
fills	O
from	O
RAM	B-Architecture
,	O
and	O
so	O
the	O
algorithm	O
contains	O
much	O
complex	O
logic	O
to	O
avoid	O
cache	B-General_Concept
misses	O
as	O
often	O
as	O
possible	O
.	O
</s>
<s>
Due	O
to	O
these	O
cache	B-General_Concept
optimizations	O
,	O
Judy	B-Data_Structure
arrays	I-Data_Structure
are	O
fast	O
,	O
especially	O
for	O
very	O
large	O
datasets	O
.	O
</s>
<s>
On	O
data	O
sets	O
that	O
are	O
sequential	O
or	O
nearly	O
sequential	O
,	O
Judy	B-Data_Structure
arrays	I-Data_Structure
can	O
even	O
outperform	O
hash	B-Algorithm
tables	I-Algorithm
,	O
since	O
,	O
unlike	O
hash	B-Algorithm
tables	I-Algorithm
,	O
the	O
internal	O
tree	O
structure	O
of	O
Judy	B-Data_Structure
arrays	I-Data_Structure
maintains	O
the	O
ordering	O
of	O
the	O
keys	O
.	O
</s>
<s>
Judy	B-Data_Structure
arrays	I-Data_Structure
are	O
extremely	O
complicated	O
.	O
</s>
<s>
In	O
addition	O
,	O
Judy	B-Data_Structure
arrays	I-Data_Structure
are	O
optimized	O
for	O
machines	O
with	O
64	O
byte	O
cache	B-General_Concept
lines	O
,	O
making	O
them	O
essentially	O
unportable	O
without	O
a	O
significant	O
rewrite	O
.	O
</s>
