ClientUtils escape query
ClientUtils.escapeQueryChars seems a bit aggressive to me in terms of what it escapes. It references http://lucene.apache.org/java/docs/queryparsersyntax.html#Escaping Special Characters, but doesn't explicitly escape them, instead opting for the more general \W regex. Thus, I'm noticing that chars that don't need to be escaped ( like / ) are being escaped. Anyone recall why this is? I suppose the problem comes in when one considers other query parsers, but maybe we should just mark this one as explicitly for use w/ the Lucene QP? -Grant
Re: ClientUtils escape query
That came after I spent a week increasing the list of things that need escaped one at a time (waiting for errors along the way...) Erik suggested I look at how the ruby client handles it... and I haven't seen any problem since them. Is there any problem with over escaping? I know it makes some things look funny. Perhaps there is a regex that will do any non-letter except ryan On Aug 5, 2008, at 8:28 AM, Grant Ingersoll wrote: ClientUtils.escapeQueryChars seems a bit aggressive to me in terms of what it escapes. It references http://lucene.apache.org/java/docs/queryparsersyntax.html#Escaping Special Characters, but doesn't explicitly escape them, instead opting for the more general \W regex. Thus, I'm noticing that chars that don't need to be escaped ( like / ) are being escaped. Anyone recall why this is? I suppose the problem comes in when one considers other query parsers, but maybe we should just mark this one as explicitly for use w/ the Lucene QP? -Grant
Re: ClientUtils escape query
In the PHP client I used these:
/**
* Escape a value for special query characters such as ':', '(',
')', '*', '?', etc.
*
* NOTE: inside a phrase fewer characters need escaped, use [EMAIL PROTECTED]
Apache_Solr_Service::escapePhrase()} instead
*
* @param string $value
* @return string
*/
static public function escape($value)
{
//list taken from http://lucene.apache.org/java/docs/
queryparsersyntax.html#Escaping%20Special%20Characters
$pattern =
'/(\+|-|&&|\|\||!|\(|\)|\{|}|\[|]|\^|"|~|\*|\?|:|\\\)/';
$replace = '\\\$1';
return preg_replace($pattern, $replace, $value);
}
/**
* Escape a value meant to be contained in a phrase for special
query characters
*
* @param string $value
* @return string
*/
static public function escapePhrase($value)
{
$pattern = '/("|\\\)/';
$replace = '\\\$1';
return preg_replace($pattern, $replace, $value);
}
helpful?
On Aug 5, 2008, at 4:16 PM, Ryan McKinley wrote:
That came after I spent a week increasing the list of things that
need escaped one at a time (waiting for errors along the way...)
Erik suggested I look at how the ruby client handles it... and I
haven't seen any problem since them.
Is there any problem with over escaping? I know it makes some
things look funny. Perhaps there is a regex that will do any non-
letter except
ryan
On Aug 5, 2008, at 8:28 AM, Grant Ingersoll wrote:
ClientUtils.escapeQueryChars seems a bit aggressive to me in terms
of what it escapes. It references http://lucene.apache.org/java/
docs/queryparsersyntax.html#Escaping Special Characters, but
doesn't explicitly escape them, instead opting for the more
general \W regex. Thus, I'm noticing that chars that don't need
to be escaped ( like / ) are being escaped.
Anyone recall why this is? I suppose the problem comes in when
one considers other query parsers, but maybe we should just mark
this one as explicitly for use w/ the Lucene QP?
-Grant
Re: ClientUtils escape query
It's mainly a problem when one wants to display the thing later, I guess. -Grant On Aug 5, 2008, at 4:16 PM, Ryan McKinley wrote: That came after I spent a week increasing the list of things that need escaped one at a time (waiting for errors along the way...) Erik suggested I look at how the ruby client handles it... and I haven't seen any problem since them. Is there any problem with over escaping? I know it makes some things look funny. Perhaps there is a regex that will do any non- letter except ryan On Aug 5, 2008, at 8:28 AM, Grant Ingersoll wrote: ClientUtils.escapeQueryChars seems a bit aggressive to me in terms of what it escapes. It references http://lucene.apache.org/java/docs/queryparsersyntax.html#Escaping Special Characters, but doesn't explicitly escape them, instead opting for the more general \W regex. Thus, I'm noticing that chars that don't need to be escaped ( like / ) are being escaped. Anyone recall why this is? I suppose the problem comes in when one considers other query parsers, but maybe we should just mark this one as explicitly for use w/ the Lucene QP? -Grant
Re: ClientUtils escape query
Wouldn't you want to reverse all escaping in that case anyway? -Mike On 5-Aug-08, at 1:45 PM, Grant Ingersoll wrote: It's mainly a problem when one wants to display the thing later, I guess. -Grant On Aug 5, 2008, at 4:16 PM, Ryan McKinley wrote: That came after I spent a week increasing the list of things that need escaped one at a time (waiting for errors along the way...) Erik suggested I look at how the ruby client handles it... and I haven't seen any problem since them. Is there any problem with over escaping? I know it makes some things look funny. Perhaps there is a regex that will do any non- letter except ryan On Aug 5, 2008, at 8:28 AM, Grant Ingersoll wrote: ClientUtils.escapeQueryChars seems a bit aggressive to me in terms of what it escapes. It references http://lucene.apache.org/java/docs/queryparsersyntax.html#Escaping Special Characters, but doesn't explicitly escape them, instead opting for the more general \W regex. Thus, I'm noticing that chars that don't need to be escaped ( like / ) are being escaped. Anyone recall why this is? I suppose the problem comes in when one considers other query parsers, but maybe we should just mark this one as explicitly for use w/ the Lucene QP? -Grant
