MultiAutoCompleteTextView token definition

Question

I have a MultiAutoCompleteTextView that's tied to a TextWatcher and an adapter that returns stuff from a database based on what's typed in the field. This is all working properly, but I'm looking for a way to change the default token definition behavior, so when a user taps back somewhere in a token and adds a character, the entire token is considered as the query input, not the text from the left-most token to the cursor.

Here's a slightly more concrete example:

1. I have data as follows: a, b, aa, ab, aab, abb, aba, bb
2. I have the following tokens entered: aa, bb
3. I put the cursor in between the two a's (i.e., a|a)
4. I enter the letter 'b'
5. I get the following suggestion: ab
6. I expected a suggestion for: aba

I hope I explained that well enough. I did a small sample project using no fancy stuffs, only defaults and a simple String array and I have the same behavior, which seems consistent with how the docs define tokens with relation to the cursor's position.

So, is there a reliable way to include suggestions returned the other way described in my example? Even our SO tag system works like that.

Answer 1

You need to set a custom MultiAutoCompleteTextView.Tokenizer on your MultiAutoCompleteTextView using the setTokenizer() function. Override the findTokenStart and findTokenEnd functions to make sure that it's grabbing the entire text being modified.