Label Search Displays too many files, need quote marks to display correctly

[Don Wallbaum]

Gentlefolk,

Confluence 5.8.17

I created a label: es-meetingnotes

I used this label with the macro "Content Report Table" which was instructed to search for the label: es-meetingnotes

The macro was instructed to display three entries. As expected the "Find more results" displayed.

When I click on the "Find more results" link, instead of seeing the expected six entries, I get 163 results.

It appears that this labelText: (es-meetingnotes) search was searching for:

es

not

es-meetingnotes

I found that when I put "es-meetingnotes" in quotes, however, the search performs correctly, displaying just the six files with the es-meetingnotes label:

 labelText: ("es-meetingnotes")

Atlassian support site states that using a dash is legal:

https://answers.atlassian.com/questions/38054097/what-are-the-best-practices-for-how-to-name-labels

Any idea what is going on? It isn't a big deal to remove a dash, I am just trying to figure out why the dash is allowing the search function to ignore the rest of the search term.

Thx!

[Stephen Deutsch · Accepted solution]

To understand this, you have to understand how Confluence search works. Confluence uses Lucene in the background, which often breaks the content into words (refered to as tokens). By using this tokenization, Lucene can find multiple words together (in the page content, for example) even if there are one or more words in the middle.

If you look at this page: https://confluence.atlassian.com/doc/confluence-search-fields-161188.html#ConfluenceSearchFields-Pages

You can see that LabelText is also tokenized. The problem comes in that Lucene treats the dash character as a token separator, much like a space, so that is why it matches both es, meetingnotes, and es-meetingnotes. That is why you need to surround it with quotes in order to find an exact match.

[Don Wallbaum]

Thank you, Stephen. Does this mean that underscores also causes this tokenization?

Are there characters then that do not result in this tokenization? If a dash tokenizes, then does an underscore? An asterisk? Are there special  characters are are acceptable to Confluence labels and don't tokenize the labelText Search function?

My confusion - and at this point irk - is that the available help on labels states that both dashes and underscores are legal, when in fact, they really are not.

https://answers.atlassian.com/questions/38054097/what-are-the-best-practices-for-how-to-name-labels

You have helped clarify an issue, thank you again!

Don