Adding more trigrams makes post query worse about codesearch

[GoogleCodeExporter]

postReader init() was not initializing restrict correctly, so all r.next() 
calls returned the full list list and not the restricted set.

Patch attached.

Original comment by dgryski on 29 Feb 2012 at 3:43

• Added labels: ****
• Removed labels: ****

Attachments:

• restrict.patch

[GoogleCodeExporter]

With this fix, the 599 files matching 'hello world' in linux-3.1.3 example ( 
http://swtch.com/~rsc/regexp/regexp4.html ) is reduced to 32:

/tmp dgryski$ CSEARCHINDEX=/tmp/.csearch-linux csearch -verbose -i -c 'hello 
world' 
2012/03/07 11:45:20 query: ("HEL"|"HEl"|"HeL"|"Hel"|"hEL"|"hEl"|"heL"|"hel") 
("ELL"|"ELl"|"ElL"|"Ell"|"eLL"|"eLl"|"elL"|"ell") 
("LLO"|"LLo"|"LlO"|"Llo"|"lLO"|"lLo"|"llO"|"llo") ("LO "|"Lo "|"lO "|"lo ") ("O 
W"|"O w"|"o W"|"o w") (" WO"|" Wo"|" wO"|" wo") 
("WOR"|"WOr"|"WoR"|"Wor"|"wOR"|"wOr"|"woR"|"wor") 
("ORL"|"ORl"|"OrL"|"Orl"|"oRL"|"oRl"|"orL"|"orl") 
("RLD"|"RLd"|"RlD"|"Rld"|"rLD"|"rLd"|"rlD"|"rld")
2012/03/07 11:45:20 post query identified 32 possible files
/tmp/linux-3.1.3/Documentation/filesystems/ramfs-rootfs-initramfs.txt: 3
/tmp/linux-3.1.3/Documentation/java.txt: 1
/tmp/linux-3.1.3/Documentation/s390/Debugging390.txt: 3
/tmp/linux-3.1.3/arch/blackfin/kernel/kgdb_test.c: 1
/tmp/linux-3.1.3/arch/frv/kernel/gdb-stub.c: 1
/tmp/linux-3.1.3/arch/mn10300/kernel/gdb-stub.c: 1
/tmp/linux-3.1.3/arch/powerpc/platforms/powermac/udbg_scc.c: 1
/tmp/linux-3.1.3/drivers/media/video/msp3400-driver.c: 1
/tmp/linux-3.1.3/drivers/net/sfc/selftest.c: 1
/tmp/linux-3.1.3/samples/kdb/kdb_hello.c: 2

Original comment by dgryski on 7 Mar 2012 at 10:48

• Added labels: ****
• Removed labels: ****

[GoogleCodeExporter]

This issue was closed by revision dbeef1612f35.

Original comment by dgryski on 2 May 2012 at 8:02

• Changed state: Fixed
• Added labels: ****
• Removed labels: ****