[LUCENE-2201] more performance improvements for snowball

• Type: Improvement
• Status: Closed
• Priority: Minor
• Resolution: Fixed
• Fix Version/s: 4.0-ALPHA
• Component/s: modules/analysis
• Lucene Fields: New, Patch Available
• Assignee: Robert Muir
• Reporter: Robert Muir
• Created: 09/Jan/10 18:55
• Updated: 28/Aug/22 12:18
• Resolved: 13/Jan/10 22:29

Description

i took a more serious look at snowball after LUCENE-2194.

This gives greatly improved performance, but note it has some minor breaks to snowball internals:

• Among.s becomes a char[] instead of a string
• SnowballProgram.current becomes a char[] instead of a StringBuilder
• SnowballProgram.eq_s(int, String) becomes eq_s(int, CharSequence), so that eq_v(StringBuilder) doesnt need to create an extra string.
• same as the above with eq_s_b and eq_v_b
• replace_s(int, int, String) becomes replace_s(int, int, CharSequence), so that StringBuilder-based slice and insertion methods don't need to create an extra string.

all of these "breaks" imho are only theoretical, the problem is just that pretty much everything is public or protected in the snowball internals.

the performance improvement here depends heavily upon the snowball language in use, but its way more significant than LUCENE-2194.