UiPath Studio

Fuzzy Search

About Fuzzy Search

A selector contains a top-level tag which references the target window or browser tab. It holds attributes like title, process name, URL, window name, and others, which help you properly identify the elements you want to use in your automation projects.

Often enough, top-level tag values are subject to dynamic changes because the target application rewrites them based on your actions. For example, a web page title can change during the automation process, which makes corresponding selector invalid.

In this regard, the selector’s fuzzy search capabilities enable you locate strings based on a pattern, rather than on an exact match to your input. This results in a more flexible method of identifying elements and validates selectors according to multiple attributes, even if some are prone to change during the automation process.

How Fuzzy Search Works

In order to use the fuzzy search capabilities, a couple of matching options need to be included in the top-level tag of the target selector, as follows:

Option
Description
matching:<tag_name>=fuzzy

Allows you to specify the target search tag, determined by the <tag_name> value.

fuzzyLevel:<tag_name>=<numerical_value>

Determines the matching threshold for the target <tag_name>. It supports a numerical value from 0 to 1, as follows:

  • Closer to 0 - returns results with lower similarity to the <tag_name>
  • Closer to 1 - returns results with higher similarity to the <tag_name>

Note:

If more elements are found, the target action is executed on the first one.

Currently, the fuzzy search is supported by the following selector types:

Selector Type
Support

<wnd>

check_mark

<html>

check_mark

<webctrl>

check_mark

<java>

no_right

<ctrl>

no_right

<sap>

no_right

<silverlight>

no_right

Tags which identify a process name.
Example: <html app="firefox.exe">

no_right

Examples of Using Fuzzy Search

Fuzzy Search can be used to identify nearly any type of attribute in the top-level tags of selectors. In the following examples, we aim to build valid selectors with attributes which is incorrectly spelled:

A. Using a High Fuzziness Level

Let's assume we want to build a selector to identify the W3Schools Online Web Tutorials <title> attribute. For the purpose of our example, the attribute is incorrectly spelled, thus making the selector invalid.

<html app='firefox.exe' title='WEScho0ls Online Web Tutorials' /> 

To fix this, we need to introduce Fuzzy Search inside the <html> tag. Since the attribute of interest is <title>, the <tag_name> of our Fuzzy Search expression needs to have the matching:title=fuzzy form. The mistake in our <title> attribute is small, so we can use a fuzziness level closer to 1. The following selector is generated, which returns the correct <title> attribute, which, in this case is W3Schools Online Web Tutorials:

<html app='firefox.exe' title='WEScho0ls Online Web Tutorials'
      matching:title='fuzzy' fuzzylevel:title='0.8' /> 

B. Using a Low Fuzziness Level

In this second example, we need to build selectors for an <aaname> attribute. The target attribute is Colorpicker, but in our example it is incorrectly spelled Clorker. Needless to say that, with an incorrectly spelled attribute, our selector is invalid, and looks something like this:

<html title='W3Schools Online Web Tutorials' /> 
<webctrl aaname='Clorker' parentid='main' tag='IMG' /> 

The Fuzzy Search needs to be included in the <html> tag. In this case, our target <tag_name> <aaname>. Since there are more mistakes in the spelling of our target, we reduce the fuzziness level closer to 0, so that there are greater chances of accurate detection. The following selector is generated, which returns the correct <aaname> attribute, which, in this case is Colorpicker:

<html title='W3Schools Online Web Tutorials' /> 
<webctrl aaname='Colorker' parentid='main' tag='IMG' matching:aaname='fuzzy' fuzzylevel:aaname='0.4' /> 

Updated 7 days ago


Fuzzy Search


Suggested Edits are limited on API Reference Pages

You can only suggest edits to Markdown body content, but not to the API spec.