Re: periods in full text

Tech-Archive recommends: Fix windows errors by optimizing your registry



Rebecca,

That works.

In case you did not read Hilary's reply to a more recent question about
percent signs, you have another option. If the dot is important to
accuracy, change the '.' into the word 'dot' and you could search for
'12dotXTdot1234dot1*'.

Anyway, glad that you have an answer that works for you.

RLF

"ronsavi" <ronsavi@xxxxxxxxxxxxxxxxxxxxxxxxx> wrote in message
news:CFD19381-AAD3-40DC-AB72-C6DB16D683C8@xxxxxxxxxxxxxxxx
Thanks a lot, Russell -
In the end, decided to strip out the dots before creating the full text
and
then strip out the dots from the search term. It's not pretty, but
effective
enough.
Thanks again,
Rebecca

"Russell Fields" wrote:

Rebecca,

I suspect that it actually has "too many -intermediate- records" to work
with, and not that it plans to return too many records to you. Still,
SQL
Server full text indexing is great when you are asking the questions that
it
understands, but difficult when the question exceeds its intent.

In many ways, your query criteria looks more like a string search than a
full text search. If it makes sense and this is what you need to do, you
might try combining the two, something like:

select count(*)
from emu_Objects
where contains(fulltextsearch,'"12.XT.1234"')
and fulltextsearch like '%12.XT.1234.1%'

But, of course, this requires a lot more understanding of the intent of
the
query to do this combined approach. Just dropping the CONTAINS() in this
case might be easier to decipher, but may perform slower.

This is all just a For What It Is Worth,
RLF

"ronsavi" <ronsavi@xxxxxxxxxxxxxxxxxxxxxxxxx> wrote in message
news:0DE34303-F4B1-4245-A909-F8E856265F13@xxxxxxxxxxxxxxxx
Hi Russell -
I was able to update the noise.enu after downloading an Unlocker tool
to
determine what had the hold on it. The search returns more accurate
results
now.

If I search for 12.XT.1234.1, I get that record back, but I only get
that
record back.

If I search for 12.XT.1234.1*, then I get an error saying that too many
records have been returned. I can only assume this is because it is
looking
to match 1* throughout the entire index, and almost every record will
have
a
1* in it somewhere. Strange that I can query on 12.XT.1234 and it
returns
the correct set of about 400, but when I add the .1, then it pushes it
over
the edge to 'too many records.' So adding a term increases the
returned
set?
Seems odd to me. Is this just how SQL Server works?

Thanks again for your help,
Rebecca

"Russell Fields" wrote:

Rebecca,

Sorry, I don't know what that would be. I edit my noise word files
without
stopping and starting the server. Perhaps a population is running at
the
moment, but I would not have expected it to hold the file open.

RLF

"ronsavi" <ronsavi@xxxxxxxxxxxxxxxxxxxxxxxxx> wrote in message
news:D63ECC4F-11B5-43E3-AE59-6B62ACC78ACE@xxxxxxxxxxxxxxxx
Thanks so much, Russel.

I'm trying to update my noise.enu file, but it keeps saying that the
file
is
in use. Do you know if I need to shut down SQLSrvr to update the
file?
I'm
not running anything else on the box except for my Windows Explorer.

Thanks,
Rebecca

"Russell Fields" wrote:

ronsavi,

Since the . is a wordbreaker, it is ignored by full text searching.
There
is not a wordbreaker that views the . in any other way. This means
that
your query is looking for 4 words together, 12, XT, 1234, and 1.
Any
search
that appears to honor the punctuation is simply an artifact of the
data
that
was found.

The reason that 12.XT.1234.1 finds 12.XT.1234.39 is likely that
your
noise
word file says that 1 is a noise word. You may need to edit your
noise
word
list and remove words that you want to search. If you remove 1
from
your
noise word list (and doubtless 2, 3, 4, 5, 6, 7, 8, 9, 0) then you
will
be
able to search for 1. (If you decide you want no noise words, you
can
delete all words in the noise list, but you must leave at least one
blank
in
the file or include 1 word like ThisWordIsToMarkAnEmptyWordList.)

In that case, if you wanted to find 12.XT.1234.1 ,
12.XT.1234.10,
12.XT.1234.19 you would search for 12.XT.1234.1*. Note that the
wildcard
is only for the word 1, not for the whole string. (Also, fulltext
indexing
will only wildcard for trailing characters, not leading
characters.)

RLF

"ronsavi" <ronsavi@xxxxxxxxxxxxxxxxxxxxxxxxx> wrote in message
news:92BA94A5-E405-41AD-A1F9-9E7EFFE17765@xxxxxxxxxxxxxxxx
Hello -
We are using:
Windows Server 2003 SP2
SQL Server 2000 SP2

Data is stored using periods such as 12.XT.1234.1

I realize that the . is a punctuation, but how exatly does it
search
for
this when I use the following contains query:

select count(*)
from emu_Objects
where contains(fulltextsearch,'"12.XT.1234.1"')

This query returns this list of 39 records:
12.XT.1234.1
12.XT.1234.2
12.XT.1234.3
... up to
12.XT.1234.39

I'd like it to return this list of 11 records:
12.XT.1234.1
12.XT.1234.10
12.XT.1234.11
... up to
12.XT.1234.19

Is there a way to do this other than using like?
I know some people have fixed their problems by switching to the
Neutral
Word Breaker, but would that help in my case?

Thanks very much!











.



Relevant Pages

  • Re: periods in full text
    ... then strip out the dots from the search term. ... query to do this combined approach. ... Is this just how SQL Server works? ... word file says that 1 is a noise word. ...
    (microsoft.public.sqlserver.fulltext)
  • Re: periods in full text
    ... query to do this combined approach. ... Is this just how SQL Server works? ... word file says that 1 is a noise word. ...
    (microsoft.public.sqlserver.fulltext)
  • Re: Internet Newbie question
    ... xxx DOT com the server will query it but will accept it. ... >> Fell in a trap already, luckily I heard it dialing and was eventually ...
    (microsoft.public.windowsxp.security_admin)
  • Re: periods in full text
    ... I edit my noise word files without ... word file says that 1 is a noise word. ... Windows Server 2003 SP2 ... This query returns this list of 39 records: ...
    (microsoft.public.sqlserver.fulltext)
  • Re: Unable to connect to DOMAIN Controller
    ... Your first mistake was to name the local AD domain the same as the internet ... > So I was forced to rebuild the server. ... The internal network is 10 dot while my external ...
    (microsoft.public.backoffice.smallbiz2000)