Mailing List Archive: Lucene: Java-Dev

[jira] Commented: (LUCENE-1126) Simplify StandardTokenizer JFlex grammar

Index | Next | Previous | View Threaded

jira at apache

Aug 27, 2008, 3:19 AM

Post #1 of 10 (605 views)
Permalink

[jira] Commented: (LUCENE-1126) Simplify StandardTokenizer JFlex grammar

[ https://issues.apache.org/jira/browse/LUCENE-1126?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=12626052#action_12626052 ]

Michael McCandless commented on LUCENE-1126:
--------------------------------------------

This patch looks reasonable? We are replacing our own hardwired regexp for DIGIT with JFlex's :digit: which then falls back to the Character.isDigit() [equivalent] on the JVM that ran jflex.

> Simplify StandardTokenizer JFlex grammar
> ----------------------------------------
>
> Key: LUCENE-1126
> URL: https://issues.apache.org/jira/browse/LUCENE-1126
> Project: Lucene - Java
> Issue Type: Improvement
> Components: Analysis
> Affects Versions: 2.2
> Reporter: Steven Rowe
> Priority: Minor
> Fix For: 2.4
>
> Attachments: LUCENE-1126.patch
>
>
> Summary of thread entitled "Fullwidth alphanumeric characters, plus a question on Korean ranges" begun by Daniel Noll on java-user, and carried over to java-dev:
> On 01/07/2008 at 5:06 PM, Daniel Noll wrote:
> > I wish the tokeniser could just use Character.isLetter and
> > Character.isDigit instead of having to know all the ranges itself, since
> > the JRE already has all this information. Character.isLetter does
> > return true for CJK characters though, so the ranges would still come in
> > handy for determining what kind of letter they are. I don't support
> > JFlex has a way to do this...
> The DIGIT macro could be replaced by JFlex's predefined character class [:digit:], which has the same semantics as java.lang.Character.isDigit().
> Although JFlex's predefined character class [:letter:] (same semantics as java.lang.Character.isLetter()) includes CJK characters, there is a way to handle this using JFlex's regex negation syntax {{!}}. From [the JFlex documentation|http://jflex.de/manual.html]:
> bq. [T]he expression that matches everything of {{a}} not matched by {{b}} is !(!{{a}}|{{b}})
> So to exclude CJ characters from the LETTER macro:
> {code}
> LETTER = ! ( ! [:letter:] | {CJ} )
> {code}
>
> Since [:letter:] includes all of the Korean ranges, there's no reason (AFAICT) to treat them separately; unlike Chinese and Japanese characters, which are individually tokenized, the Korean characters should participate in the same token boundary rules as all of the other letters.
> I looked at some of the differences between Unicode 3.0.0, which Java 1.4.2 supports, and Unicode 5.0, the latest version, and there are lots of new and modified letter and digit ranges. This stuff gets tweaked all the time, and I don't think Lucene should be in the business of trying to track it, or take a position on which Unicode version users' data should conform to.
> Switching to using JFlex's [:letter:] and [:digit:] predefined character classes ties (most of) these decisions to the user's choice of JVM version, and this seems much more reasonable to me than the current status quo.
> I will attach a patch shortly.

--
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.

---------------------------------------------------------------------
To unsubscribe, e-mail: java-dev-unsubscribe [at] lucene
For additional commands, e-mail: java-dev-help [at] lucene

jira at apache

Aug 28, 2008, 4:53 AM

Post #2 of 10 (566 views)
Permalink

[jira] Commented: (LUCENE-1126) Simplify StandardTokenizer JFlex grammar [In reply to]

[ https://issues.apache.org/jira/browse/LUCENE-1126?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=12626531#action_12626531 ]

Michael McCandless commented on LUCENE-1126:
--------------------------------------------

I will commit this in a day or two.

> Simplify StandardTokenizer JFlex grammar
> ----------------------------------------
>
> Key: LUCENE-1126
> URL: https://issues.apache.org/jira/browse/LUCENE-1126
> Project: Lucene - Java
> Issue Type: Improvement
> Components: Analysis
> Affects Versions: 2.2
> Reporter: Steven Rowe
> Assignee: Michael McCandless
> Priority: Minor
> Fix For: 2.4
>
> Attachments: LUCENE-1126.patch
>
>
> Summary of thread entitled "Fullwidth alphanumeric characters, plus a question on Korean ranges" begun by Daniel Noll on java-user, and carried over to java-dev:
> On 01/07/2008 at 5:06 PM, Daniel Noll wrote:
> > I wish the tokeniser could just use Character.isLetter and
> > Character.isDigit instead of having to know all the ranges itself, since
> > the JRE already has all this information. Character.isLetter does
> > return true for CJK characters though, so the ranges would still come in
> > handy for determining what kind of letter they are. I don't support
> > JFlex has a way to do this...
> The DIGIT macro could be replaced by JFlex's predefined character class [:digit:], which has the same semantics as java.lang.Character.isDigit().
> Although JFlex's predefined character class [:letter:] (same semantics as java.lang.Character.isLetter()) includes CJK characters, there is a way to handle this using JFlex's regex negation syntax {{!}}. From [the JFlex documentation|http://jflex.de/manual.html]:
> bq. [T]he expression that matches everything of {{a}} not matched by {{b}} is !(!{{a}}|{{b}})
> So to exclude CJ characters from the LETTER macro:
> {code}
> LETTER = ! ( ! [:letter:] | {CJ} )
> {code}
>
> Since [:letter:] includes all of the Korean ranges, there's no reason (AFAICT) to treat them separately; unlike Chinese and Japanese characters, which are individually tokenized, the Korean characters should participate in the same token boundary rules as all of the other letters.
> I looked at some of the differences between Unicode 3.0.0, which Java 1.4.2 supports, and Unicode 5.0, the latest version, and there are lots of new and modified letter and digit ranges. This stuff gets tweaked all the time, and I don't think Lucene should be in the business of trying to track it, or take a position on which Unicode version users' data should conform to.
> Switching to using JFlex's [:letter:] and [:digit:] predefined character classes ties (most of) these decisions to the user's choice of JVM version, and this seems much more reasonable to me than the current status quo.
> I will attach a patch shortly.

--
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.

---------------------------------------------------------------------
To unsubscribe, e-mail: java-dev-unsubscribe [at] lucene
For additional commands, e-mail: java-dev-help [at] lucene

jira at apache

Sep 3, 2008, 2:27 AM

Post #3 of 10 (543 views)
Permalink

[jira] Commented: (LUCENE-1126) Simplify StandardTokenizer JFlex grammar [In reply to]

[ https://issues.apache.org/jira/browse/LUCENE-1126?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=12627943#action_12627943 ]

Michael McCandless commented on LUCENE-1126:
--------------------------------------------

Hmm -- I'm now seeing an failure with this patch, in TestThaiAnalyzer (in contrib/analyzers):

{code}
[junit] Testcase: testAnalyzer(org.apache.lucene.analysis.th.TestThaiAnalyzer): FAILED
[junit] expected:<?[]> but was:<?[??]>
[junit] junit.framework.ComparisonFailure: expected:<?[]> but was:<?[??]>
[junit] at org.apache.lucene.analysis.th.TestThaiAnalyzer.assertAnalyzesTo(TestThaiAnalyzer.java:43)
[junit] at org.apache.lucene.analysis.th.TestThaiAnalyzer.testAnalyzer(TestThaiAnalyzer.java:54)
[junit]
{code}

Does anyone else see this?

> Simplify StandardTokenizer JFlex grammar
> ----------------------------------------
>
> Key: LUCENE-1126
> URL: https://issues.apache.org/jira/browse/LUCENE-1126
> Project: Lucene - Java
> Issue Type: Improvement
> Components: Analysis
> Affects Versions: 2.2
> Reporter: Steven Rowe
> Assignee: Michael McCandless
> Priority: Minor
> Fix For: 2.4
>
> Attachments: LUCENE-1126.patch
>
>
> Summary of thread entitled "Fullwidth alphanumeric characters, plus a question on Korean ranges" begun by Daniel Noll on java-user, and carried over to java-dev:
> On 01/07/2008 at 5:06 PM, Daniel Noll wrote:
> > I wish the tokeniser could just use Character.isLetter and
> > Character.isDigit instead of having to know all the ranges itself, since
> > the JRE already has all this information. Character.isLetter does
> > return true for CJK characters though, so the ranges would still come in
> > handy for determining what kind of letter they are. I don't support
> > JFlex has a way to do this...
> The DIGIT macro could be replaced by JFlex's predefined character class [:digit:], which has the same semantics as java.lang.Character.isDigit().
> Although JFlex's predefined character class [:letter:] (same semantics as java.lang.Character.isLetter()) includes CJK characters, there is a way to handle this using JFlex's regex negation syntax {{!}}. From [the JFlex documentation|http://jflex.de/manual.html]:
> bq. [T]he expression that matches everything of {{a}} not matched by {{b}} is !(!{{a}}|{{b}})
> So to exclude CJ characters from the LETTER macro:
> {code}
> LETTER = ! ( ! [:letter:] | {CJ} )
> {code}
>
> Since [:letter:] includes all of the Korean ranges, there's no reason (AFAICT) to treat them separately; unlike Chinese and Japanese characters, which are individually tokenized, the Korean characters should participate in the same token boundary rules as all of the other letters.
> I looked at some of the differences between Unicode 3.0.0, which Java 1.4.2 supports, and Unicode 5.0, the latest version, and there are lots of new and modified letter and digit ranges. This stuff gets tweaked all the time, and I don't think Lucene should be in the business of trying to track it, or take a position on which Unicode version users' data should conform to.
> Switching to using JFlex's [:letter:] and [:digit:] predefined character classes ties (most of) these decisions to the user's choice of JVM version, and this seems much more reasonable to me than the current status quo.
> I will attach a patch shortly.

--
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.

---------------------------------------------------------------------
To unsubscribe, e-mail: java-dev-unsubscribe [at] lucene
For additional commands, e-mail: java-dev-help [at] lucene

jira at apache

Sep 3, 2008, 12:33 PM

Post #4 of 10 (544 views)
Permalink

[jira] Commented: (LUCENE-1126) Simplify StandardTokenizer JFlex grammar [In reply to]

[ https://issues.apache.org/jira/browse/LUCENE-1126?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=12628106#action_12628106 ]

Steven Rowe commented on LUCENE-1126:
-------------------------------------

Yeah, I see this too.

The issue is that the entire Thai range {{\u0e00-\u0e5b}} is included in the unpatched grammar's {LETTER} definition, which contains the huge range {{\u0100-\u1fff}}, much of which is not actually letters. The patched grammar instead substitutes the Unicode 3.0 {{Letter}} general category (via JFlex's [:letter:]), which excludes some characters in the Thai range: non-spacing marks, a currency symbol, numerals, etc.

ThaiAnalyzer uses ThaiWordFilter, which uses Java's BreakIterator to tokenize the contiguous text (i.e. without whitespace) provided by StandardTokenizer.

The failing test expects to see {{"\u0e17\u0e35\u0e48"}}, but instead gets {{"\u0e17"}}, because {{\u0e35}} is a non-spacing mark, which the patched StandardTokenizer doesn't pass to ThaiWordFilter.

Because of this problem, I guess I'm -1 on applying the patch I provided.

One solution would be to switch from using the {{Letter}} general category to the derived property {{Alphabetic}}, which includes both general categories {{Letter}} and {{Mark}}. (see Annex C of [the Unicode Regular Expressions Technical Standard|http://www.unicode.org/unicode/reports/tr18/#Compatibility_Properties] under "alpha" for discussion of this). The current version of JFlex does not support Unicode property references in its syntax, though, so simplifying -- and correcting -- the grammar may have to wait for the next version of JFlex, which will support syntax like {{\p{Alphabetic}}}.

> Simplify StandardTokenizer JFlex grammar
> ----------------------------------------
>
> Key: LUCENE-1126
> URL: https://issues.apache.org/jira/browse/LUCENE-1126
> Project: Lucene - Java
> Issue Type: Improvement
> Components: Analysis
> Affects Versions: 2.2
> Reporter: Steven Rowe
> Assignee: Michael McCandless
> Priority: Minor
> Fix For: 2.4
>
> Attachments: LUCENE-1126.patch
>
>
> Summary of thread entitled "Fullwidth alphanumeric characters, plus a question on Korean ranges" begun by Daniel Noll on java-user, and carried over to java-dev:
> On 01/07/2008 at 5:06 PM, Daniel Noll wrote:
> > I wish the tokeniser could just use Character.isLetter and
> > Character.isDigit instead of having to know all the ranges itself, since
> > the JRE already has all this information. Character.isLetter does
> > return true for CJK characters though, so the ranges would still come in
> > handy for determining what kind of letter they are. I don't support
> > JFlex has a way to do this...
> The DIGIT macro could be replaced by JFlex's predefined character class [:digit:], which has the same semantics as java.lang.Character.isDigit().
> Although JFlex's predefined character class [:letter:] (same semantics as java.lang.Character.isLetter()) includes CJK characters, there is a way to handle this using JFlex's regex negation syntax {{!}}. From [the JFlex documentation|http://jflex.de/manual.html]:
> bq. [T]he expression that matches everything of {{a}} not matched by {{b}} is !(!{{a}}|{{b}})
> So to exclude CJ characters from the LETTER macro:
> {code}
> LETTER = ! ( ! [:letter:] | {CJ} )
> {code}
>
> Since [:letter:] includes all of the Korean ranges, there's no reason (AFAICT) to treat them separately; unlike Chinese and Japanese characters, which are individually tokenized, the Korean characters should participate in the same token boundary rules as all of the other letters.
> I looked at some of the differences between Unicode 3.0.0, which Java 1.4.2 supports, and Unicode 5.0, the latest version, and there are lots of new and modified letter and digit ranges. This stuff gets tweaked all the time, and I don't think Lucene should be in the business of trying to track it, or take a position on which Unicode version users' data should conform to.
> Switching to using JFlex's [:letter:] and [:digit:] predefined character classes ties (most of) these decisions to the user's choice of JVM version, and this seems much more reasonable to me than the current status quo.
> I will attach a patch shortly.

--
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.

---------------------------------------------------------------------
To unsubscribe, e-mail: java-dev-unsubscribe [at] lucene
For additional commands, e-mail: java-dev-help [at] lucene

jira at apache

Sep 4, 2008, 4:13 AM

Post #5 of 10 (533 views)
Permalink

[jira] Commented: (LUCENE-1126) Simplify StandardTokenizer JFlex grammar [In reply to]

[ https://issues.apache.org/jira/browse/LUCENE-1126?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=12628308#action_12628308 ]

Michael McCandless commented on LUCENE-1126:
--------------------------------------------

{quote}
One solution would be to switch from using the Letter general category to the derived property Alphabetic, which includes both general categories Letter and Mark. (see Annex C of the Unicode Regular Expressions Technical Standard under "alpha" for discussion of this). The current version of JFlex does not support Unicode property references in its syntax, though, so simplifying - and correcting - the grammar may have to wait for the next version of JFlex, which will support syntax like \p{Alphabetic}.
{quote}
Could we, alternatively, modify the patch to explicitly add back in the full Thai range into ALPHANUM, and then upgrade to \p{Alphabetic} once the next version of JFlex is released?

Or are there other languages, besides Thai, that we might break with this patch?

> Simplify StandardTokenizer JFlex grammar
> ----------------------------------------
>
> Key: LUCENE-1126
> URL: https://issues.apache.org/jira/browse/LUCENE-1126
> Project: Lucene - Java
> Issue Type: Improvement
> Components: Analysis
> Affects Versions: 2.2
> Reporter: Steven Rowe
> Assignee: Michael McCandless
> Priority: Minor
> Fix For: 2.4
>
> Attachments: LUCENE-1126.patch
>
>
> Summary of thread entitled "Fullwidth alphanumeric characters, plus a question on Korean ranges" begun by Daniel Noll on java-user, and carried over to java-dev:
> On 01/07/2008 at 5:06 PM, Daniel Noll wrote:
> > I wish the tokeniser could just use Character.isLetter and
> > Character.isDigit instead of having to know all the ranges itself, since
> > the JRE already has all this information. Character.isLetter does
> > return true for CJK characters though, so the ranges would still come in
> > handy for determining what kind of letter they are. I don't support
> > JFlex has a way to do this...
> The DIGIT macro could be replaced by JFlex's predefined character class [:digit:], which has the same semantics as java.lang.Character.isDigit().
> Although JFlex's predefined character class [:letter:] (same semantics as java.lang.Character.isLetter()) includes CJK characters, there is a way to handle this using JFlex's regex negation syntax {{!}}. From [the JFlex documentation|http://jflex.de/manual.html]:
> bq. [T]he expression that matches everything of {{a}} not matched by {{b}} is !(!{{a}}|{{b}})
> So to exclude CJ characters from the LETTER macro:
> {code}
> LETTER = ! ( ! [:letter:] | {CJ} )
> {code}
>
> Since [:letter:] includes all of the Korean ranges, there's no reason (AFAICT) to treat them separately; unlike Chinese and Japanese characters, which are individually tokenized, the Korean characters should participate in the same token boundary rules as all of the other letters.
> I looked at some of the differences between Unicode 3.0.0, which Java 1.4.2 supports, and Unicode 5.0, the latest version, and there are lots of new and modified letter and digit ranges. This stuff gets tweaked all the time, and I don't think Lucene should be in the business of trying to track it, or take a position on which Unicode version users' data should conform to.
> Switching to using JFlex's [:letter:] and [:digit:] predefined character classes ties (most of) these decisions to the user's choice of JVM version, and this seems much more reasonable to me than the current status quo.
> I will attach a patch shortly.

--
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.

---------------------------------------------------------------------
To unsubscribe, e-mail: java-dev-unsubscribe [at] lucene
For additional commands, e-mail: java-dev-help [at] lucene

jira at apache

Sep 4, 2008, 9:13 AM

Post #6 of 10 (532 views)
Permalink

[jira] Commented: (LUCENE-1126) Simplify StandardTokenizer JFlex grammar [In reply to]

[ https://issues.apache.org/jira/browse/LUCENE-1126?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=12628368#action_12628368 ]

Steven Rowe commented on LUCENE-1126:
-------------------------------------

{quote}
Could we, alternatively, modify the patch to explicitly add back in the full Thai range into ALPHANUM, and then upgrade to \p{Alphabetic} once the next version of JFlex is released?

Or are there other languages, besides Thai, that we might break with this patch?
{quote}

I noticed in looking at the Unicode database that Lao, which is contiguous with Thai and contained in the unpatched {{{LETTER}}} range, has exactly the same issue as Thai. However, the Lucene code base doesn't contain a Lao Analyzer. And I think ThaiAnalyzer is depending on faulty behavior from StandardTokenizer, so "fixing" the issue for other languages would be to make the assumption that they too would depend on bad behavior.

I'll shortly provide a redone patch that allows the ThaiAnalyzer to work again, but unless we have evidence of actual use by other language analyzers, I don't think we should be addressing them right now.

> Simplify StandardTokenizer JFlex grammar
> ----------------------------------------
>
> Key: LUCENE-1126
> URL: https://issues.apache.org/jira/browse/LUCENE-1126
> Project: Lucene - Java
> Issue Type: Improvement
> Components: Analysis
> Affects Versions: 2.2
> Reporter: Steven Rowe
> Assignee: Michael McCandless
> Priority: Minor
> Fix For: 2.4
>
> Attachments: LUCENE-1126.patch
>
>
> Summary of thread entitled "Fullwidth alphanumeric characters, plus a question on Korean ranges" begun by Daniel Noll on java-user, and carried over to java-dev:
> On 01/07/2008 at 5:06 PM, Daniel Noll wrote:
> > I wish the tokeniser could just use Character.isLetter and
> > Character.isDigit instead of having to know all the ranges itself, since
> > the JRE already has all this information. Character.isLetter does
> > return true for CJK characters though, so the ranges would still come in
> > handy for determining what kind of letter they are. I don't support
> > JFlex has a way to do this...
> The DIGIT macro could be replaced by JFlex's predefined character class [:digit:], which has the same semantics as java.lang.Character.isDigit().
> Although JFlex's predefined character class [:letter:] (same semantics as java.lang.Character.isLetter()) includes CJK characters, there is a way to handle this using JFlex's regex negation syntax {{!}}. From [the JFlex documentation|http://jflex.de/manual.html]:
> bq. [T]he expression that matches everything of {{a}} not matched by {{b}} is !(!{{a}}|{{b}})
> So to exclude CJ characters from the LETTER macro:
> {code}
> LETTER = ! ( ! [:letter:] | {CJ} )
> {code}
>
> Since [:letter:] includes all of the Korean ranges, there's no reason (AFAICT) to treat them separately; unlike Chinese and Japanese characters, which are individually tokenized, the Korean characters should participate in the same token boundary rules as all of the other letters.
> I looked at some of the differences between Unicode 3.0.0, which Java 1.4.2 supports, and Unicode 5.0, the latest version, and there are lots of new and modified letter and digit ranges. This stuff gets tweaked all the time, and I don't think Lucene should be in the business of trying to track it, or take a position on which Unicode version users' data should conform to.
> Switching to using JFlex's [:letter:] and [:digit:] predefined character classes ties (most of) these decisions to the user's choice of JVM version, and this seems much more reasonable to me than the current status quo.
> I will attach a patch shortly.

--
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.

---------------------------------------------------------------------
To unsubscribe, e-mail: java-dev-unsubscribe [at] lucene
For additional commands, e-mail: java-dev-help [at] lucene

jira at apache

Sep 4, 2008, 10:07 AM

Post #7 of 10 (531 views)
Permalink

[jira] Commented: (LUCENE-1126) Simplify StandardTokenizer JFlex grammar [In reply to]

[ https://issues.apache.org/jira/browse/LUCENE-1126?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=12628387#action_12628387 ]

Michael McCandless commented on LUCENE-1126:
--------------------------------------------

bq. I'll shortly provide a redone patch that allows the ThaiAnalyzer to work again, but unless we have evidence of actual use by other language analyzers, I don't think we should be addressing them right now.

OK this sounds like the right approach.

> Simplify StandardTokenizer JFlex grammar
> ----------------------------------------
>
> Key: LUCENE-1126
> URL: https://issues.apache.org/jira/browse/LUCENE-1126
> Project: Lucene - Java
> Issue Type: Improvement
> Components: Analysis
> Affects Versions: 2.2
> Reporter: Steven Rowe
> Assignee: Michael McCandless
> Priority: Minor
> Fix For: 2.4
>
> Attachments: LUCENE-1126.patch
>
>
> Summary of thread entitled "Fullwidth alphanumeric characters, plus a question on Korean ranges" begun by Daniel Noll on java-user, and carried over to java-dev:
> On 01/07/2008 at 5:06 PM, Daniel Noll wrote:
> > I wish the tokeniser could just use Character.isLetter and
> > Character.isDigit instead of having to know all the ranges itself, since
> > the JRE already has all this information. Character.isLetter does
> > return true for CJK characters though, so the ranges would still come in
> > handy for determining what kind of letter they are. I don't support
> > JFlex has a way to do this...
> The DIGIT macro could be replaced by JFlex's predefined character class [:digit:], which has the same semantics as java.lang.Character.isDigit().
> Although JFlex's predefined character class [:letter:] (same semantics as java.lang.Character.isLetter()) includes CJK characters, there is a way to handle this using JFlex's regex negation syntax {{!}}. From [the JFlex documentation|http://jflex.de/manual.html]:
> bq. [T]he expression that matches everything of {{a}} not matched by {{b}} is !(!{{a}}|{{b}})
> So to exclude CJ characters from the LETTER macro:
> {code}
> LETTER = ! ( ! [:letter:] | {CJ} )
> {code}
>
> Since [:letter:] includes all of the Korean ranges, there's no reason (AFAICT) to treat them separately; unlike Chinese and Japanese characters, which are individually tokenized, the Korean characters should participate in the same token boundary rules as all of the other letters.
> I looked at some of the differences between Unicode 3.0.0, which Java 1.4.2 supports, and Unicode 5.0, the latest version, and there are lots of new and modified letter and digit ranges. This stuff gets tweaked all the time, and I don't think Lucene should be in the business of trying to track it, or take a position on which Unicode version users' data should conform to.
> Switching to using JFlex's [:letter:] and [:digit:] predefined character classes ties (most of) these decisions to the user's choice of JVM version, and this seems much more reasonable to me than the current status quo.
> I will attach a patch shortly.

--
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.

---------------------------------------------------------------------
To unsubscribe, e-mail: java-dev-unsubscribe [at] lucene
For additional commands, e-mail: java-dev-help [at] lucene

jira at apache

Sep 4, 2008, 10:55 AM

Post #8 of 10 (532 views)
Permalink

[jira] Commented: (LUCENE-1126) Simplify StandardTokenizer JFlex grammar [In reply to]

[ https://issues.apache.org/jira/browse/LUCENE-1126?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=12628406#action_12628406 ]

Michael McCandless commented on LUCENE-1126:
--------------------------------------------

Steven, it looks like you ran JFlex with a 1.5 or 1.6 JRE? Should we stick with that, or, go with a 1.4 JRE (which is indeed significantly different)?

> Simplify StandardTokenizer JFlex grammar
> ----------------------------------------
>
> Key: LUCENE-1126
> URL: https://issues.apache.org/jira/browse/LUCENE-1126
> Project: Lucene - Java
> Issue Type: Improvement
> Components: Analysis
> Affects Versions: 2.2
> Reporter: Steven Rowe
> Assignee: Michael McCandless
> Priority: Minor
> Fix For: 2.4
>
> Attachments: LUCENE-1126.patch, LUCENE-1126.patch
>
>
> Summary of thread entitled "Fullwidth alphanumeric characters, plus a question on Korean ranges" begun by Daniel Noll on java-user, and carried over to java-dev:
> On 01/07/2008 at 5:06 PM, Daniel Noll wrote:
> > I wish the tokeniser could just use Character.isLetter and
> > Character.isDigit instead of having to know all the ranges itself, since
> > the JRE already has all this information. Character.isLetter does
> > return true for CJK characters though, so the ranges would still come in
> > handy for determining what kind of letter they are. I don't support
> > JFlex has a way to do this...
> The DIGIT macro could be replaced by JFlex's predefined character class [:digit:], which has the same semantics as java.lang.Character.isDigit().
> Although JFlex's predefined character class [:letter:] (same semantics as java.lang.Character.isLetter()) includes CJK characters, there is a way to handle this using JFlex's regex negation syntax {{!}}. From [the JFlex documentation|http://jflex.de/manual.html]:
> bq. [T]he expression that matches everything of {{a}} not matched by {{b}} is !(!{{a}}|{{b}})
> So to exclude CJ characters from the LETTER macro:
> {code}
> LETTER = ! ( ! [:letter:] | {CJ} )
> {code}
>
> Since [:letter:] includes all of the Korean ranges, there's no reason (AFAICT) to treat them separately; unlike Chinese and Japanese characters, which are individually tokenized, the Korean characters should participate in the same token boundary rules as all of the other letters.
> I looked at some of the differences between Unicode 3.0.0, which Java 1.4.2 supports, and Unicode 5.0, the latest version, and there are lots of new and modified letter and digit ranges. This stuff gets tweaked all the time, and I don't think Lucene should be in the business of trying to track it, or take a position on which Unicode version users' data should conform to.
> Switching to using JFlex's [:letter:] and [:digit:] predefined character classes ties (most of) these decisions to the user's choice of JVM version, and this seems much more reasonable to me than the current status quo.
> I will attach a patch shortly.

--
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.

---------------------------------------------------------------------
To unsubscribe, e-mail: java-dev-unsubscribe [at] lucene
For additional commands, e-mail: java-dev-help [at] lucene

jira at apache

Sep 4, 2008, 11:33 AM

Post #9 of 10 (530 views)
Permalink

[jira] Commented: (LUCENE-1126) Simplify StandardTokenizer JFlex grammar [In reply to]

[ https://issues.apache.org/jira/browse/LUCENE-1126?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=12628424#action_12628424 ]

Steven Rowe commented on LUCENE-1126:
-------------------------------------

bq. Steven, it looks like you ran JFlex with a 1.5 or 1.6 JRE? Should we stick with that, or, go with a 1.4 JRE (which is indeed significantly different)?

You're right, Mike - I was inadvertently using a 1.5 JRE. I'll put up another patch produced with 1.4 JRE shortly - we should definitely not go with 1.5..

> Simplify StandardTokenizer JFlex grammar
> ----------------------------------------
>
> Key: LUCENE-1126
> URL: https://issues.apache.org/jira/browse/LUCENE-1126
> Project: Lucene - Java
> Issue Type: Improvement
> Components: Analysis
> Affects Versions: 2.2
> Reporter: Steven Rowe
> Assignee: Michael McCandless
> Priority: Minor
> Fix For: 2.4
>
> Attachments: LUCENE-1126.patch, LUCENE-1126.patch
>
>
> Summary of thread entitled "Fullwidth alphanumeric characters, plus a question on Korean ranges" begun by Daniel Noll on java-user, and carried over to java-dev:
> On 01/07/2008 at 5:06 PM, Daniel Noll wrote:
> > I wish the tokeniser could just use Character.isLetter and
> > Character.isDigit instead of having to know all the ranges itself, since
> > the JRE already has all this information. Character.isLetter does
> > return true for CJK characters though, so the ranges would still come in
> > handy for determining what kind of letter they are. I don't support
> > JFlex has a way to do this...
> The DIGIT macro could be replaced by JFlex's predefined character class [:digit:], which has the same semantics as java.lang.Character.isDigit().
> Although JFlex's predefined character class [:letter:] (same semantics as java.lang.Character.isLetter()) includes CJK characters, there is a way to handle this using JFlex's regex negation syntax {{!}}. From [the JFlex documentation|http://jflex.de/manual.html]:
> bq. [T]he expression that matches everything of {{a}} not matched by {{b}} is !(!{{a}}|{{b}})
> So to exclude CJ characters from the LETTER macro:
> {code}
> LETTER = ! ( ! [:letter:] | {CJ} )
> {code}
>
> Since [:letter:] includes all of the Korean ranges, there's no reason (AFAICT) to treat them separately; unlike Chinese and Japanese characters, which are individually tokenized, the Korean characters should participate in the same token boundary rules as all of the other letters.
> I looked at some of the differences between Unicode 3.0.0, which Java 1.4.2 supports, and Unicode 5.0, the latest version, and there are lots of new and modified letter and digit ranges. This stuff gets tweaked all the time, and I don't think Lucene should be in the business of trying to track it, or take a position on which Unicode version users' data should conform to.
> Switching to using JFlex's [:letter:] and [:digit:] predefined character classes ties (most of) these decisions to the user's choice of JVM version, and this seems much more reasonable to me than the current status quo.
> I will attach a patch shortly.

--
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.

---------------------------------------------------------------------
To unsubscribe, e-mail: java-dev-unsubscribe [at] lucene
For additional commands, e-mail: java-dev-help [at] lucene

jira at apache

Sep 4, 2008, 11:41 AM

Post #10 of 10 (536 views)
Permalink

[jira] Commented: (LUCENE-1126) Simplify StandardTokenizer JFlex grammar [In reply to]

[ https://issues.apache.org/jira/browse/LUCENE-1126?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=12628428#action_12628428 ]

Michael McCandless commented on LUCENE-1126:
--------------------------------------------

Steven, it's OK I can regen; I just wanted to confirm which JRE (1.4) we should use.

I'm also going to add a comment at the top of StandardTokenizerImpl.jflex stating this.

> Simplify StandardTokenizer JFlex grammar
> ----------------------------------------
>
> Key: LUCENE-1126
> URL: https://issues.apache.org/jira/browse/LUCENE-1126
> Project: Lucene - Java
> Issue Type: Improvement
> Components: Analysis
> Affects Versions: 2.2
> Reporter: Steven Rowe
> Assignee: Michael McCandless
> Priority: Minor
> Fix For: 2.4
>
> Attachments: LUCENE-1126.patch, LUCENE-1126.patch
>
>
> Summary of thread entitled "Fullwidth alphanumeric characters, plus a question on Korean ranges" begun by Daniel Noll on java-user, and carried over to java-dev:
> On 01/07/2008 at 5:06 PM, Daniel Noll wrote:
> > I wish the tokeniser could just use Character.isLetter and
> > Character.isDigit instead of having to know all the ranges itself, since
> > the JRE already has all this information. Character.isLetter does
> > return true for CJK characters though, so the ranges would still come in
> > handy for determining what kind of letter they are. I don't support
> > JFlex has a way to do this...
> The DIGIT macro could be replaced by JFlex's predefined character class [:digit:], which has the same semantics as java.lang.Character.isDigit().
> Although JFlex's predefined character class [:letter:] (same semantics as java.lang.Character.isLetter()) includes CJK characters, there is a way to handle this using JFlex's regex negation syntax {{!}}. From [the JFlex documentation|http://jflex.de/manual.html]:
> bq. [T]he expression that matches everything of {{a}} not matched by {{b}} is !(!{{a}}|{{b}})
> So to exclude CJ characters from the LETTER macro:
> {code}
> LETTER = ! ( ! [:letter:] | {CJ} )
> {code}
>
> Since [:letter:] includes all of the Korean ranges, there's no reason (AFAICT) to treat them separately; unlike Chinese and Japanese characters, which are individually tokenized, the Korean characters should participate in the same token boundary rules as all of the other letters.
> I looked at some of the differences between Unicode 3.0.0, which Java 1.4.2 supports, and Unicode 5.0, the latest version, and there are lots of new and modified letter and digit ranges. This stuff gets tweaked all the time, and I don't think Lucene should be in the business of trying to track it, or take a position on which Unicode version users' data should conform to.
> Switching to using JFlex's [:letter:] and [:digit:] predefined character classes ties (most of) these decisions to the user's choice of JVM version, and this seems much more reasonable to me than the current status quo.
> I will attach a patch shortly.

--
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.

---------------------------------------------------------------------
To unsubscribe, e-mail: java-dev-unsubscribe [at] lucene
For additional commands, e-mail: java-dev-help [at] lucene

Index | Next | Previous | View Threaded

Interested in having your list archived? Contact Gossamer Threads