TWiki Formatted Search

Inline search feature allows flexible formatting of search result

The default output format of a %SEARCH{...}% is a table consisting of topic names and topic summaries. Use the format="..." parameter to customize the search result. The format parameter typically defines a bullet or a table row containing variables, such as %SEARCH{ "food" format="| $topic | $summary |" }%. See %SEARCH{...}% for other search parameters, such as separator="".

On this page:

Syntax

Two parameters can be used to specify a customized search result:

1. header="..." parameter

Use the header parameter to specify the header of a search result. It should correspond to the format of the format parameter. This parameter is optional.
Example: header="| *Topic:* | *Summary:* |"

Variables that can be used in the header string:

Name: Expands To:
$web Name of the web
$n or $n() New line. Use $n() if followed by alphanumeric character, e.g. write Foo$n()Bar instead of Foo$nBar
$nop or $nop() Is a "no operation". This variable gets removed; useful for nested search
$quot Double quote (") (\" also works)
$percnt Percent sign (%)
$dollar Dollar sign ($)

2. format="..." parameter

Use the format parameter to specify the format of one search hit.
Example: format="| $topic | $summary |"

Variables that can be used in the format string:

Name: Expands To:
$web Name of the web
$topic Topic name
$topic(20) Topic name, "- " hyphenated each 20 characters
$topic(30, -<br />) Topic name, hyphenated each 30 characters with separator "-<br />"
$topic(40, ...) Topic name, shortended to 40 characters with "..." indication
$parent Name of parent topic; empty if not set
$parent(20) Name of parent topic, same hyphenation/shortening like $topic()
$text Formatted topic text. In case of a multiple="on" search, it is the line found for each search hit.
$locked LOCKED flag (if any)
$date Time stamp of last topic update, e.g. 29 Jul 2010 - 15:04
$isodate Time stamp of last topic update, e.g. 2010-07-29T15:04Z
$rev Number of last topic revision, e.g. 4
$username Login name of last topic update, e.g. jsmith
$wikiname Wiki user name of last topic update, e.g. JohnSmith
$wikiusername Wiki user name of last topic update, like Main.JohnSmith
$createdate Time stamp of topic revision 1
$createusername Login name of topic revision 1, e.g. jsmith
$createwikiname Wiki user name of topic revision 1, e.g. JohnSmith
$createwikiusername Wiki user name of topic revision 1, e.g. Main.JohnSmith
$summary Topic summary, just the plain text, all formatting and line breaks removed; up to 162 characters
$summary(50) Topic summary, up to 50 characters shown
$summary(showvarnames) Topic summary, with %ALLTWIKI{...}% variables shown as ALLTWIKI{...}
$summary(noheader) Topic summary, with leading ---+ headers removed
Note: The tokens can be combined, for example $summary(100, showvarnames, noheader)
$changes Summary of changes between latest rev and previous rev
$changes(n) Summary of changes between latest rev and rev n
$formname The name of the form attached to the topic; empty if none
$formfield(name) The field value of a form field; for example, $formfield(TopicClassification) would get expanded to PublicFAQ. This applies only to topics that have a TWikiForm
$formfield(name, 10) Form field value, "- " hyphenated each 10 characters
$formfield(name, 20, -<br />) Form field value, hyphenated each 20 characters with separator "-<br />"
$formfield(name, 30, ...) Form field value, shortended to 30 characters with "..." indication
$pattern(reg-exp) A regular expression pattern to extract some text from a topic (does not search meta data; use $formfield instead). In case of a multiple="on" search, the pattern is applied to the line found in each search hit.
• Specify a Regular Expression that covers the whole text (topic or line), which typically starts with .*, and must end in .*
• Put text you want to keep in parenthesis, like $pattern(.*?(from here.*?to here).*)
• Example: $pattern(.*?\*.*?Email\:\s*([^\n\r]+).*) extracts the e-mail address from a bullet of format * Email: ...
• This example has non-greedy .*? patterns to scan for the first occurance of the Email bullet; use greedy .* patterns to scan for the last occurance
• Limitation: Do not use .*) inside the pattern, e.g. $pattern(.*foo(.*)bar.*) does not work, but $pattern(.*foo(.*?)bar.*) does
• Note: Make sure that the integrity of a web page is not compromised; for example, if you include an HTML table make sure to include everything including the table end tag
$count(reg-exp) Count of number of times a regular expression pattern appears in the text of a topic (does not search meta data). Follows guidelines for use and limitations outlined above under $pattern(reg-exp). Example: $count(.*?(---[+][+][+][+]) .*) counts the number of <H4> headers in a page.
$n or $n() New line. Use $n() if followed by alphanumeric character, e.g. write Foo$n()Bar instead of Foo$nBar
$nop or $nop() Is a "no operation". This variable gets removed; useful for nested search
$quot Double quote (") (\" also works)
$percnt Percent sign (%)
$dollar Dollar sign ($)

Examples

Here are some samples of formatted searches. The Search Pattern Cookbook has other examples, such as creating a picklist of usernames, searching for topic children and more.

Bullet list showing topic name and summary

Write this:

%SEARCH{ "FAQ" scope="topic" nosearch="on" nototal="on" header="   * *Topic: Summary:*" format="   * [[$topic]]: $summary" }%

To get this:

Table showing form field values of topics with a form

In a web where there is a form that contains a TopicClassification field, an OperatingSystem field and an OsVersion field we could write:

| *Topic:* | *OperatingSystem:* | *OsVersion:* |
%SEARCH{ "[T]opicClassification.*?value=\"[P]ublicFAQ\"" scope="text" type="regex" nosearch="on" nototal="on" format="| [[$topic]] | $formfield(OperatingSystem) | $formfield(OsVersion) |" }%

To get this:

Topic: OperatingSystem: OsVersion:
IncorrectDllVersionW32PTH10DLL OsWin 95/98
WinDoze95Crash OsWin 95

Extract some text from a topic using regular expression

Write this:

%SEARCH{ "__Back to\:__ TWikiFAQ" scope="text" type="regex" nosearch="on" nototal="on" header="TWiki FAQs:" format="   * $pattern(.*?FAQ\:[\n\r]*([^\n\r]+).*) [[$topic][Answer...]]" }%

To get this:

TWiki FAQs:

Nested Search

Search can be nested. For example, search for some topics, then form a new search for each topic found in the first search. The idea is to build the nested search string using a formatted search in the first search.

Here is an example. Let's search for all topics that contain the word "culture" (first search), and let's find out where each topic found is linked from (second search).

Write this:

%SEARCH{ "culture" format="   * $topic is referenced by:$n      * $percntSEARCH{ \"$topic\" format=\"$dollartopic\" nosearch=\"on\" nototal=\"on\" separator=\", \" }$nop%" nosearch="on" nototal="on" }%

To get this:

Note: Nested search can be slow, especially if you nest more then 3 times. Nesting is limited to 16 levels. For each new nesting level you need to "escape the escapes", e.g. write $dollarpercntSEARCH{ for level three, $dollardollarpercntSEARCH{ for level four, etc.

Most recently changed pages

Write this:

%SEARCH{ "\.*" scope="topic" type="regex" nosearch="on" nototal="on" order="modified" reverse="on"  format="| [[$topic]] | $wikiusername  | $date |" limit="7" }%

To get this:

TWiki Preferences Geri Ragghianti 09 Feb 2009 - 03:25
Web Left Bar Login Geri Ragghianti 09 Feb 2009 - 02:54
Web Left Bar Example Geri Ragghianti 09 Feb 2009 - 02:54
Web Left Bar Cookbook Geri Ragghianti 09 Feb 2009 - 02:54
Pattern Skin Css Geri Ragghianti 09 Feb 2009 - 02:54
New User Template Geri Ragghianti 09 Feb 2009 - 02:44
Web Top Bar Login Geri Ragghianti 09 Feb 2009 - 02:10

Search with conditional output

A regular expression search is flexible, but there are limitations. For example, you cannot show all topics that are up to exactly one week old, or create a report that shows all records with invalid form fields or fields within a certain range, etc. You need some additional logic to format output based on a condition:

  1. Specify a search which returns more hits then you need
  2. For each search hit apply a spreadsheet formula to determine if the hit is needed
  3. If needed, format and output the result
  4. Else supress the search hit

This requires the TWiki:Plugins.SpreadSheetPlugin. The following example shows all topics that are up to exactly one week old.

Write this:

%CALC{$SET(weekold, $TIMEADD($TIME(), -7, day))}%
%SEARCH{ "." scope="topic" type="regex" nosearch="on" nototal="on" order="modified" reverse="on" format="$percntCALC{$IF($TIME($date) < $GET(weekold), <nop>, | [[$topic]] | $wikiusername | $date | $rev |)}$percnt" limit="100" }%

To get this:

Embedding search forms to return a formatted result

Use an HTML form and an embedded formatted search on the same topic. You can link them together with an %URLPARAM{"..."}% variable. Example:

Write this: 

<form action="%SCRIPTURLPATH{"view"}%/%WEB%/%TOPIC%">
Find Topics: 
<input type="text" name="q" size="32" value="%URLPARAM{"q"}%" />&nbsp;<input type="submit" class="twikiSubmit" value="Search" />
</form>
Result:
%SEARCH{ search="%URLPARAM{"q"}%" format="   * $web.$topic: %BR% $summary" nosearch="on" }%

To get this:

Find Topics:  
Result:

Related Topics: User Documentation Category, Search Help, TWiki Variables#Var SEARCH, Search Pattern Cookbook, Regular Expression

-- Contributors: TWiki:Main.PeterThoeny, TWiki:Main.CrawfordCurrie



Topic FormattedSearch . { Edit | Attach | Backlinks: Web All webs | Printable | History: r19 < r18 < r17 < r16 < r15 More }