Bugzilla

Here's how I tested on a Mac:
1. Download a Hebrew version of Nightly
2. Enable `showSearchTerms.featureGate`
3. Switch my keyboard to Hebrew
4. Type either a set of non-numeric Hebrew characters or Hebrew characters with a number
5. Do a search

While using a mixture of text and numbers, I found the search term matches the ordering of characters in a search results page.

I found it's only when the search consists entirely of numbers that the term in the urlbar doesn't match.

After some digging, it's because in SearchEngine, we call `Services.textToSubURI.ConvertAndEscape(this.queryCharset, data)` which gives us a string that isn't reversed if the `data` consisting of only numbers.

Here's how I tested on a Mac:
1. Download a Hebrew version of Nightly
2. Enable `showSearchTerms.featureGate`
3. Switch my keyboard to Hebrew
4. Type either a set of non-numeric Hebrew characters or Hebrew characters with a number
5. Do a search

While using a mixture of text and numbers, I found the search term matches the ordering of characters in a search results page.

I found it's only when the search consists entirely of numbers that the term in the urlbar doesn't match.

After some digging, it's because in SearchEngine, we call `Services.textToSubURI.ConvertAndEscape(this.queryCharset, data)` which gives us a string that isn't reversed if the `data` consists solely of numbers. 

That makes sense because the function can't infer that the string should be reversed without giving it additional context, namely characters from a particular bidirectional language.

Here's how I tested on a Mac:
1. Download a Hebrew version of Nightly
2. Enable `showSearchTerms.featureGate`
3. Switch my keyboard to Hebrew
4. Type either a set of non-numeric Hebrew characters or Hebrew characters with a number
5. Do a search

While using a mixture of text and numbers, I found the search term matches the ordering of characters in a search results page.

When the search consists entirely of numbers, the term in the urlbar won't match.

After some digging, it's because in SearchEngine, we call `Services.textToSubURI.ConvertAndEscape(this.queryCharset, data)`. It does a conversion of the characters into a UTF-8 string. The difference is we encode a pure number string in the same order as they are input since there's nothing to suggest otherwise, whereas the the search engines encode the numbers in reverse.