[jsword-devel] False search hits with certain locales
DM Smith
dmsmith at crosswire.org
Wed Feb 8 13:44:44 MST 2012
The simplest fix (though not right in an av11n context) is to change
KeyUtil.getPassage(Key)
from
try {
ref = keyf.getKey(key.getName());
} catch (NoSuchKeyException ex) {
log.warn("Key can't be a passage: " + key.getName());
ref = keyf.createEmptyKeyList();
}
to
try {
ref = keyf.getKey(key.getOsisID());
} catch (NoSuchKeyException ex) {
log.warn("Key can't be a passage: " + key.getName());
ref = keyf.createEmptyKeyList();
}
On 02/08/2012 03:14 PM, DM Smith wrote:
> On 02/08/2012 03:05 PM, Martin Denham wrote:
>> I found the problem:
>> Rev.Full = Johannes\u2019 openberring
>>
>> \u0219 is an apostrophe and so it was matching Johannes' openberring
>> 22:8 but stopping at the apostrophe which of course matched the whole
>> of John.
>>
>> Best regards
>> Martin
>
> That's part of it. I just looked at it and the bigger, other part is
> that KeyUtil.getPassage(Key) tries to cast a Key to a passage. It
> should not be called on a Verse or a VerseRange. As it gets the locale
> version of the Verse and tries to convert that to a Passage.
>
> It didn't need to convert the verse to Norwegian, and then re-parse
> it, incorrectly into a passage. It had everything it needed in the Verse.
>
> I've got to think about that for a bit to figure out the best way to
> fix it and where. Fixing it will be a performance improvement in general.
>
> The apostrophe in the name will cause other problems in JSword.
> Likewise for other "punctuation". But that is another problem.
>
> In Him,
> DM
>
>>
>>
>> On 8 February 2012 19:23, Martin Denham <mjdenham at gmail.com
>> <mailto:mjdenham at gmail.com>> wrote:
>>
>> I have just noticed that I have not fixed the problem. I am now
>> getting an error on the final hit 'Key can't be a passage' - I
>> don't know what that means:
>> 02-08 19:05:34.105: I/System.out(22191): 129 found:Johannes'
>> openberring 1:1 docid=30681 docbase=0 key.card:1 res.card=129
>> 02-08 19:05:34.105: I/System.out(22191): 130 found:Johannes'
>> openberring 1:4 docid=30684 docbase=0 key.card:1 res.card=130
>> 02-08 19:05:34.105: I/System.out(22191): 131 found:Johannes'
>> openberring 1:9 docid=30689 docbase=0 key.card:1 res.card=131
>> 02-08 19:05:34.145: I/System.out(22191): JSword:Key can't be a
>> passage: Johannes' openberring 22:8
>> 02-08 19:05:34.155: I/System.out(22191): 132 found:Johannes'
>> openberring 22:8 docid=31071 docbase=0 key.card:1 res.card=131
>>
>> To log the cardinality I just added a println in the
>> VerseCollector as below:
>> Key key =
>> VerseFactory.fromString(doc.get(LuceneIndex.FIELD_KEY));
>> results.addAll(key);
>> System.out.println(++count + " found:"
>> +key.getName()+ " docid="+docId+" docbase="+docBase+"
>> key.card:"+key.getCardinality()+"
>> res.card="+results.getCardinality());
>>
>> The problem is I can't see the bug on Windows, only when running
>> on my Android phone, so I am not sure anybody without an Android
>> will be able to reproduce the problem easily.
>>
>> Martin
>>
>> On 8 February 2012 19:04, DM Smith <dmsmith at crosswire.org
>> <mailto:dmsmith at crosswire.org>> wrote:
>>
>> I've been trying to get to it, but haven't be able to do so.
>> I'd be interested in your code to log the cardinality.
>> -- DM
>>
>>
>> On 02/08/2012 01:54 PM, Martin Denham wrote:
>>> I don't know what is going on but I have done more analysis
>>> and found a fix for Nynorsk, but I think it is affecting
>>> other locales like Japanese which I can't explain.
>>>
>>> Test: search for 'John' in NT in And Bible with locale set to nn
>>> Result: 1389 hits including every verse in the gospel of John
>>> Observation: I logged the cardinality of the results var in
>>> VerseCollector and you can see that it jumps from 131 to
>>> 1389 on the last hit in Rev.22.8:
>>> 02-08 18:18:15.895: I/System.out(21945): 127
>>> found:Apostelgjerningane 19:4 docid=27575 docbase=0
>>> key.card:1 res.card=127
>>> 02-08 18:18:15.905: I/System.out(21945): 128
>>> found:Galatarane 2:9 docid=29073 docbase=0 key.card:1
>>> res.card=128
>>> 02-08 18:18:15.905: I/System.out(21945): 129 found:Johannes'
>>> openberring 1:1 docid=30681 docbase=0 key.card:1 res.card=129
>>> 02-08 18:18:15.915: I/System.out(21945): 130 found:Johannes'
>>> openberring 1:4 docid=30684 docbase=0 key.card:1 res.card=130
>>> 02-08 18:18:15.915: I/System.out(21945): 131 found:Johannes'
>>> openberring 1:9 docid=30689 docbase=0 key.card:1 res.card=131
>>> 02-08 18:18:15.965: I/System.out(21945): 132 found:Johannes'
>>> openberring 22:8 docid=31071 docbase=0 key.card:1 res.card=1389
>>>
>>> Other words in Rev 22 seem to have the same effect e.g.
>>> month, behold, am,...
>>>
>>> The fix for nn was to change
>>> Rev.Short=Op
>>> to
>>> Rev.Short=JoOp
>>>
>>> Any idea what is happening? I tried to write a junit on my
>>> pc but couldn't get it to fail on Windows.
>>>
>>> I am using revision 2195 of JSword, which is before the AV
>>> changes.
>>>
>>> Thanks
>>> Martin
>>>
>>>
>>> On 2 February 2012 11:20, DM Smith <dmsmith at crosswire.org
>>> <mailto:dmsmith at crosswire.org>> wrote:
>>>
>>> I'm trying to see what is happening. It doesn't make
>>> sense to me either.
>>>
>>> Cent from my fone so theer mite be tipos. ;)
>>>
>>> On Jan 27, 2012, at 9:44 AM, Martin Denham
>>> <mjdenham at gmail.com <mailto:mjdenham at gmail.com>> wrote:
>>>
>>>> Hi,
>>>>
>>>> I have received this error report for And Bible
>>>> <http://code.google.com/p/and-bible/issues/detail?id=87> which
>>>> has confused me. I would be grateful for any
>>>> suggestions wrt what might be happening.
>>>>
>>>> A simple test I have tried:
>>>>
>>>> * Set locale to de or en
>>>> * Search for 'John' in ESV
>>>> * Works fine
>>>> * Set locale to nn (Norsk Nynorsk)
>>>> * Search for 'John' in ESV
>>>> * Every verse of John is returned in the result list
>>>>
>>>> Thanks
>>>> Martin
>>>
>
>
>
> _______________________________________________
> jsword-devel mailing list
> jsword-devel at crosswire.org
> http://www.crosswire.org/mailman/listinfo/jsword-devel
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://www.crosswire.org/pipermail/jsword-devel/attachments/20120208/d4b19335/attachment.html>
More information about the jsword-devel
mailing list