Please fix the search function

[Cthulhu D]

I want to be able to search for a phrase like "opening balanced 11 counts"

 

I cannot because the tool won't let me search when there is a word that has less than 4 letters. This is fine as a normal restriction because of the database impact, but when I am searching for a complete *phrase* that is completely ridiculous - I'm effectively searching for one much longer string. It prohibits so many useful searches

 

"Opening hearts and clubs"

 

"When should you open 1NT with an unbalanced hand"

 

Please let the entire string be considered for the 4 characters or less constraint, not just each sub string.

[TylerE]

You're much better off doing something like this, really:

 

http://lmgtfy.com/?q=opening+balanced+11+counts+site%3Abridgebase.com%2Fforums

[inquiry]

TylerE reply is basically right... here it is in a little more detail

 

Go to http://www.google.com/advanced_search

 

enter http://www.bridgebase.com/forums/ as the domain to search

 

 

 

 

Enter the phrase of words you want to search for.... if you enter the phrase "opening balanced 11 counts" you get three threads, including yours (This one).... leave out the quotes you get a lot more relevant ones on the first few pages.

 

 

 

 

Why can't you search for short words? The software that runs the forum does not allow search for short words, (three letters I think). We change that, so use google...

 

 

 

[Vampyr]

Why can't you search for short words? The software that runs the forum does not allow search for short words, (three letters I think). We change that, so use google...

 

It's very strange, isn't it. But you say you can change it? If that is not what you meant, whoever maintains the software could, and should. I don't know about this database thing, but if there is an issue there it should just be made possible to search for a phrase In quotation marks, or a single word surrounded by spaces and in quotes.

[Antrax]

You can't search for short words to prevent a malicious user from deliberately making searches that return huge sets of results, straining the server the forums are on.

It's sufficient to add the string "site:bridgebase.com" (no quotation marks needed) to normal Google queries to limit the search to things in the forums.

[Cthulhu D]

You can't search for short words to prevent a malicious user from deliberately making searches that return huge sets of results, straining the server the forums are on.

 

I understand that, but in the case of much longer strings such as "Opening balanced 11 counts" the string has 26 characters in it and thus will not crush the database.

 

You're much better off doing something like this, really:

 

http://lmgtfy.com/?q=opening+balanced+11+counts+site%3Abridgebase.com%2Fforums

 

That has some key limitations you'll notice:

 

A) You'll notice due to how google presents results if you ran the same search on both - such as unbalanced diamond opening systems - you actually get fairly different results. For example: jgillispie's thread about Magic diamond systems (right in the wheelhouse) doesn't show up for the first 6 pages of google results for me. I stopped checking.

 

B) Please explain how I limit the results to a particular sub forum using google like I can with the forum software? It doesn't work

 

C) Please explain how I find all posts by Frances Hinden on the topic of opening balanced 11 counts? Again, not a feature that is offered

 

Seriously I know site:XXXX works but it's not nearly as useful as having the native search tool properly configured. The inability to search for phrases is totally hamstringing the native search for literally no reason.

[helene_t]

Long time ago, in 1995, when harddisks were more expensive than they are today, I worked as a software engineer for a data warehouse that provided the search engines for about a thousand public libraries. We excluded certain very common words from the indexing because if you want to search the phrase "See you in a day or two", the engine would first have to retrieve links to all items containing the word "see", the same for "you" etc, merge all those enormouse lists together, and finally filter out those in which the word order was not as desired. OK it was slightly more efficient than that but you get the picture.

 

This made it impossible to find books with a title made of only common words, so we made second index in which book titles were treated as single words. This had the limitation that you still couldn't find the book if you only knew that it contained the sequence "in a day or" but at least you could find it if you knew the first few words, or, alternatively, the last few words.

 

Google has some very clever engineers and enormous computer resources. But I think that for a small database like the BBF it should not be a problem even with sligtly suboptimal search engine and limited computer resources. I am obviously wrong for some reason :)

[mycroft]

by "can", I think from when I asked this question, inquiry meant "can't".

 

Yes, it's a real issue: "1NT" "IMP", "XX", ... I think my last one was "I want to find where I said that X of their 1NT was one-suited, forcing, but I was going to pass". After taking out X, 1NT, and one-, I ended up searching all my posts - it was faster.

 

ltgtfy seems to be the way to go, but I keep forgetting about it.

[barmar]

by "can", I think from when I asked this question, inquiry meant "can't".

The minimum word length is an option in forum administration, but we've chosen to keep it at its default of 4 letters. If you lower it, the index size can really explode, due to extremely common words like "the" and "if".

 

Since you can do these searches using Google, there doesn't seem to be an overriding need for us to adjust our settings.

[TylerE]

The minimum word length is an option in forum administration, but we've chosen to keep it at its default of 4 letters. If you lower it, the index size can really explode, due to extremely common words like "the" and "if".

 

Since you can do these searches using Google, there doesn't seem to be an overriding need for us to adjust our settings.

 

Is it possible to white-list certain shorter sequences, e.g. "XX" or "1NT" while not lowering the general limit?r

[Antrax]

Or alternatively, lower the limit but blacklist common words like "and" and "the"?

[Cthulhu D]

I'm not making myself clear. The functionality in the search tool where if you put a set of search terms in quotation marks it searches for that exact phrase is not excluded from the prevent searches with less than 4 characters in them, when it should be because it has a lot more than four characters. For example:

 

"1NT one suited and forcing"

should be a perfectly valid search as it is *one string* in the same way that

"Magic Diamond"

is.

 

All I want is to be able to search for phrases properly. That will solve the issue and not require white lists or anything else.

[helene_t]

Cthulhu, it is not that simple. What gets indexed is words only. "1NT one suited and forcing" won't be in the index because the user might as well search on "one suited and" or any other subphrase so the index would be too big. For the same reason, the word "diamond" can't be found using the query "iamo" unless you grep through all the documents which would be too slow.

 

Searching on a sentence works as I described it. You look up the individual words in the index. Then you merge the search results, i.e. if there is a hit for "1NT" as the 38th word in some post it looks for "one" hits in the same post and filter out those that are not in the 39th position etc.

 

The way we handled such a query was first to mask "one" and "and" because they are too common (the search result list would be too big). So we would look for

"1NT" in the nth position

"suited" in the n+2the position

"forcing" in the n+4th position

Finally we would retrieve the identified documents and verify that the original phrase was exactly as asked for, including "one" and "and".

 

At the time I left the company they were working on more clever algorithms that identified the most characteristic feature of a query such a a propper noun or even an unusual character sequence within a word, so that the result list could be narrowed down quickly. This would not work well for end users because a typo would often be the most "characteristic" feature of a query, but our software was made for librarians who rarely make typos in their queries.

[barmar]

Is it possible to white-list certain shorter sequences, e.g. "XX" or "1NT" while not lowering the general limit?r

No, there's no whitelist.

 

The search function uses MySQL's Full-Text Search feature, described here: http://dev.mysql.com/doc/refman/5.0/en/fulltext-search.html

 

I did make a mistake earlier, though. It has a blacklist of common words (called "stop list" in the documentation), so even lowering the word length won't cause it to index words like "the". So it wouldn't be as bad as I thought to lower the limit.

 

But since you can do these searches with Google, and I'll bet it does them better, I don't see this as a priority.

[Vampyr]

 

But since you can do these searches with Google, and I'll bet it does them better, I don't see this as a priority.

 

Post #6 gives a few reasons why it is not "better".

[mgoetze]

Since you can do these searches using Google, there doesn't seem to be an overriding need for us to adjust our settings.

 

B) Please explain how I limit the results to a particular sub forum using google like I can with the forum software? It doesn't work

 

C) Please explain how I find all posts by Frances Hinden on the topic of opening balanced 11 counts? Again, not a feature that is offered

[barmar]

mgoetze, if I could have answered Cthulhu's questions when they posted them, I would have.