[sword-devel] Sword phrase search not returning all expected results

Sat Mar 1 03:45:09 EST 2025

Hi Tobias,

A Lucene search for 'faith' in the KJV module using Xiphos returns 231 locations.
Aside: Only 2 of these locations are in the OT !!!

I'm sure there are verses that have the word repeated, as the whole word occurs 247 times.
(Search results from the plain text file output by diatheke)

If one drops the whole word criterion, the total leaps to 362, as words such as 'faithful', 'faithfulness', 'faithless', etc., are then included.

Best regards,

David

Sent with [Proton Mail](https://proton.me/mail/home) secure email.

On Saturday, March 1st, 2025 at 7:09 AM, Tobias Klein <contact at tklein.info> wrote:

> Hi Troy,
>
> can this be fixed in SWORD?
>
> This bug impacts the search function quite significantly. I noticed when my standard test scenario for search started to fail after my adjustments.
> The reason was that the search results for my test scenario significantly increased and I had to adjust the expected results.
> The test scenario searches for "faith" in KJV. Previously (before the bugfix) I expected 324 search results.
> After the bugfix/change mentioned below there are now 338 search results. So you see that quite some verses are missed by the search function because of this bug.
>
> Best regards,
> Tobias
>
> On 2/23/25 18:38, David Haslam wrote:
>
>> Excellent sleuthing, Tobias !
>>
>> Best regards,
>>
>> David
>>
>> Sent with [Proton Mail](https://proton.me/mail/home) secure email.
>>
>> On Sunday, February 23rd, 2025 at 5:17 PM, Tobias Klein [<contact at tklein.info>](mailto:contact at tklein.info) wrote:
>>
>>> Hi Troy,
>>>
>>> I have discovered the root cause of this bug.
>>>
>>> There is the following code in osisplain.cpp.
>>> I suppose the uppercasing action here has negative impact on the overall parsing when the stripText() is running?
>>>
>>> elseif (!strncmp(token, "/divineName", 11)) {
>>> // Get the end portion of the string, and upper case it
>>> char*end=buf.getRawData();
>>> end+=buf.size() -u->lastTextNode.size();
>>> toupperstr(end);
>>> }
>>> When I comment this portion out, the search bug does not occur anymore and I get a correct result, see below.
>>>
>>> textBuf: For he said, Because the Lord hath sworn that the Lord will have war with Amalek from generation to generation.
>>> term: generation to generation
>>> Got 11 results!
>>> Exod 17:16
>>> Isa 13:20
>>> Isa 34:10
>>> Isa 34:17
>>> Isa 51:8
>>> Jer 50:39
>>> Lam 5:19
>>> Dan 4:3
>>> Dan 4:34
>>> Joel 3:20
>>> Luke 1:50
>>>
>>> So, what the code stumbles over in the specific case of Exodus 17:16 is the <divineName> tag and the parsing / actions related to it.
>>> Why is the uppercasing necessary at all in the code above? Shouldn't this be left to the application software in terms of formatting the respective element/tag in uppercase?
>>>
>>> Best regards,
>>> Tobias
>>>
>>> On 2/22/25 20:32, Tobias Klein wrote:
>>>
>>>> Hi Troy,
>>>>
>>>> so I did a little debugging on this.
>>>>
>>>> The respective portion of code in swmodule.cpp is this code below. I added some conditional print outs for Exodus 17:16 to see what happens there.
>>>>
>>>> caseSEARCHTYPE_PHRASE: {
>>>> textBuf=stripText();
>>>> if ((flags&REG_ICASE) ==REG_ICASE) textBuf.toUpper();
>>>> SWKey*currentKey=getKey();
>>>> std::stringreferenceKey=["Exod 17:16"](http://Exod17:16);
>>>> if (currentKey->getShortText() ==referenceKey) {
>>>> std::cout<<"textBuf: "<<textBuf.c_str() <<std::endl;
>>>> std::cout<<"term: "<<term.c_str() <<std::endl;
>>>> }
>>>> // TKL: This is where the actual search per verse happens
>>>> sres=strstr(textBuf.c_str(), term.c_str());
>>>>
>>>> I get the following output based on my modification above:
>>>>
>>>> textBuf: For he said, Because the
>>>> term: generation to generation
>>>>
>>>> The full verse content of Exodus 17:16 in KJV is this:
>>>> For he said, Because the Lord hath sworn that the Lord will have war with Amalek from generation to generation.
>>>>
>>>> So ... it seems that the stripText() call strips too much of the content (textBuf) of the verse away.
>>>> Based on that there is no way for the strstr call to succeed detecting the term "generation to generation", because at that point it is not part of the search string (textBuf) anymore.
>>>>
>>>> Could you do some investigation regarding the behavior of stripText here?
>>>>
>>>> Best regards,
>>>> Tobias
>>>>
>>>> On 2/22/25 15:45, Tobias Klein wrote:
>>>>
>>>>> Hi Troy,
>>>>>
>>>>> an Ezra Bible App user reported that the phrase search is not working as expected.
>>>>>
>>>>> Here is an example where the results are not as expected.
>>>>>
>>>>> Module: KJV
>>>>>
>>>>> Search term: "generation to generation"
>>>>>
>>>>> I get the following results from the SWORD engine:
>>>>> Isa 13:20
>>>>> Isa 34:10
>>>>> Isa 34:17
>>>>> Isa 51:8
>>>>> Jer 50:39
>>>>> Dan 4:3
>>>>> Dan 4:34
>>>>> Joel 3:20
>>>>> Luke 1:50
>>>>>
>>>>> However, the verse Exodus 17:16 also contains this phrase, but is not in the list of search results.
>>>>> Could it be related to the way how the markup is structured?
>>>>>
>>>>> In Exodus 17:16 [KJV], the markup of the respective phrase looks like this:
>>>>>
>>>>> <w class="strong:H01755">from generation</w> <w class="strong:H01755">to generation</w>
>>>>>
>>>>> This is how I call the search function of the SWORD engine:
>>>>> listKey = module->search(searchTerm.c_str(), int(searchType), flags, scope, 0, internalModuleSearchProgressCB);
>>>>> see https://github.com/ezra-bible-app/node-sword-interface/blob/master/src/sword_backend/module_search.cpp#L178
>>>>>
>>>>> Have a nice weekend!
>>>>>
>>>>> Best regards,
>>>>> Tobias
>>>>>
>>>>> _______________________________________________
>>>>> sword-devel mailing list: sword-devel at crosswire.org
>>>>> http://crosswire.org/mailman/listinfo/sword-devel
>>>>> Instructions to unsubscribe/change your settings at above page
>>>>
>>>> _______________________________________________
>>>> sword-devel mailing list:
>>>> sword-devel at crosswire.org
>>>>
>>>> http://crosswire.org/mailman/listinfo/sword-devel
>>>> Instructions to unsubscribe/change your settings at above page
>>
>> _______________________________________________
>> sword-devel mailing list:
>> sword-devel at crosswire.org
>>
>> http://crosswire.org/mailman/listinfo/sword-devel
>> Instructions to unsubscribe/change your settings at above page
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://crosswire.org/pipermail/sword-devel/attachments/20250301/b58c4153/attachment-0001.htm>