Thunderstone Support Forums

We have recently added punctuation to our list of indexed characters. My engineering department tells me that this is the indexing call that we are using:

tsql -q "set keepnoise='on';set delexp=0;set addexp='\alnum{1,99}';set addexp='>>\alpha{1,50},=\alpha{1,50}';set addexp='[\alnum\punct]{1,30}';set addexp='\punct{1,5}';create metamorph inverted index idxmtblnew_CLEANUP on tblnew(CLEANUP);"

What is the syntax necessary to search for punctuation characters that may have other meaning in Texis? For instance, how do you search for the following: *, ", ', -, (, ), etc.? (For instance, it appears that you must enter an asterisk [*] twice in order to literally search for the asterisk character.)

Here is the complete list of those characters we want to be able to search for:

! Exclamation mark
¡ Inverted exclamation mark
? Question mark
# Pound Sign
- Dash
> Greater Than
< Less Than
; Semicolon
: Colon
( Right Parenthesis
) Left Parenthesis
[ Left Bracket
] Right Bracket
{ Left Curly Bracket
} Right Curly Bracket
' Apostrophe
. Period
$ Dollar Sign
@ At sign
& Ampersand
* Asterisk
" Double Quote

By the way, I am a technical writer here at CaseCentral, and I have been asked to research this issue while one of our key programmers is out of the office. Let me know if there is a better way for me to interact with Thunderstone to get this information. Thanks.

- jswartz@casecentral.com

The Metamorph language used in the like/likep query has several reserved prefixes:
/ # % @ w/ ( ~
They only have meaning at the beginning of a term. There are different ways to search for those prefixes literally. The simplest is to put a backslash in front of those characters
\@home
\w/sheet
\(abc)
\~mine
\#24

Sentence enders period, comma, and question are always stripped from the end of the query. You can put the term in parens or place a noise term that will get stripped as the last term (assuming you haven't turned on keepnoise).
(termination.)
termination. a

* is a wildcard anywhere in a term. To remove the special meaning of * you would have to put the term in a paren list with another non-matching term like
(abc*,qjxz)

Comma and hyphen will separate terms just like space
abc,def
is the same as
abc-def
is the same as
abc def
To force it to be treated single term put quotes around the whole term
"abc,def"
"abc-def"

Hyphen continues to have meaning at search time unless you
<sql "set hyphenphrase=0"></sql>
first and include the hyphenated term in a paren list with another non-matching term like
(abc-def,qjxz)

Yes, I believe we are setting hyphenphrase=0.

So, for hyphens and asterisks, you have to include them in a paren list with *another non-matching term*? Can you not simply search for the following:

(abc*)

Or this:

(abc-def)

Also, are there special syntax rules in searching for any of the following:

! Exclamation mark
¡ Inverted exclamation mark
? Question mark
> Greater Than
< Less Than
; Semicolon
: Colon
( Right Parenthesis
) Left Parenthesis
[ Left Bracket
] Right Bracket
{ Left Curly Bracket
} Right Curly Bracket
. Period
$ Dollar Sign
& Ampersand
" Double Quote

(abc*) won't work. Since it's a one item list it will be treated as abc* .

. , ? ( were addressed above.

For " use \" .

The rest should work as-is.

Are you certain that the following should not work "as is":

( Right Parenthesis
) Left Parenthesis
" Double Quote

I would think that you would need to use some special syntax for these?

Please reread the previous message.

Can you tell me what specific characters are considered "punct"? Are they the following:

! " # % & ' ( ) * + , - . / : ; < = > ? [ \ ]
^ { | } ~ $

What about right and left apostrophes and double quotes?

Thanks.

The definition of punct varies with the operating system, but will generally include everything that displays and is not a space. It generally is limited to the ASCII range, which does include the back-quote `, but does not include the left and right single and double quotes ‘ ’ “ ”