Fix the max-depth variable in spider
parent
720d634388
commit
74f3579644
|
@ -8,9 +8,23 @@ continuations calendar prettyprint dlists deques locals
|
||||||
spider.unique-deque combinators concurrency.semaphores ;
|
spider.unique-deque combinators concurrency.semaphores ;
|
||||||
IN: spider
|
IN: spider
|
||||||
|
|
||||||
TUPLE: spider base count max-count sleep max-depth initial-links
|
TUPLE: spider
|
||||||
filters spidered todo nonmatching quiet currently-spidering
|
base
|
||||||
#threads semaphore follow-robots? robots ;
|
{ count integer initial: 0 }
|
||||||
|
{ max-count number initial: 1/0. }
|
||||||
|
sleep
|
||||||
|
{ max-depth integer initial: 0 }
|
||||||
|
initial-links
|
||||||
|
filters
|
||||||
|
spidered
|
||||||
|
todo
|
||||||
|
nonmatching
|
||||||
|
quiet?
|
||||||
|
currently-spidering
|
||||||
|
{ #threads integer initial: 1 }
|
||||||
|
semaphore
|
||||||
|
follow-robots?
|
||||||
|
robots ;
|
||||||
|
|
||||||
TUPLE: spider-result url depth headers
|
TUPLE: spider-result url depth headers
|
||||||
fetched-in parsed-html links processed-in fetched-at ;
|
fetched-in parsed-html links processed-in fetched-at ;
|
||||||
|
@ -22,11 +36,8 @@ fetched-in parsed-html links processed-in fetched-at ;
|
||||||
over >>currently-spidering
|
over >>currently-spidering
|
||||||
swap 0 <unique-deque> [ push-url ] keep >>todo
|
swap 0 <unique-deque> [ push-url ] keep >>todo
|
||||||
<unique-deque> >>nonmatching
|
<unique-deque> >>nonmatching
|
||||||
0 >>max-depth
|
|
||||||
0 >>count
|
|
||||||
1/0. >>max-count
|
|
||||||
H{ } clone >>spidered
|
H{ } clone >>spidered
|
||||||
1 [ >>#threads ] [ <semaphore> >>semaphore ] bi ;
|
1 <semaphore> >>semaphore ;
|
||||||
|
|
||||||
: <spider-result> ( url depth -- spider-result )
|
: <spider-result> ( url depth -- spider-result )
|
||||||
spider-result new
|
spider-result new
|
||||||
|
@ -91,9 +102,9 @@ fetched-in parsed-html links processed-in fetched-at ;
|
||||||
now >>fetched-at drop ;
|
now >>fetched-at drop ;
|
||||||
|
|
||||||
:: spider-page ( spider spider-result -- )
|
:: spider-page ( spider spider-result -- )
|
||||||
spider quiet>> [ spider-result print-spidering ] unless
|
spider quiet?>> [ spider-result print-spidering ] unless
|
||||||
spider spider-result fill-spidered-result
|
spider spider-result fill-spidered-result
|
||||||
spider quiet>> [ spider-result describe ] unless
|
spider quiet?>> [ spider-result describe ] unless
|
||||||
spider spider-result add-spidered ;
|
spider spider-result add-spidered ;
|
||||||
|
|
||||||
\ spider-page ERROR add-error-logging
|
\ spider-page ERROR add-error-logging
|
||||||
|
@ -108,7 +119,7 @@ fetched-in parsed-html links processed-in fetched-at ;
|
||||||
: spider-page? ( spider -- ? )
|
: spider-page? ( spider -- ? )
|
||||||
{
|
{
|
||||||
[ todo>> deque>> deque-empty? not ]
|
[ todo>> deque>> deque-empty? not ]
|
||||||
[ [ todo>> peek-url depth>> ] [ max-depth>> ] bi < ]
|
[ [ todo>> peek-url depth>> ] [ max-depth>> ] bi <= ]
|
||||||
[ [ count>> ] [ max-count>> ] bi < ]
|
[ [ count>> ] [ max-count>> ] bi < ]
|
||||||
} 1&& ;
|
} 1&& ;
|
||||||
|
|
||||||
|
|
Loading…
Reference in New Issue