Performance and usability improvements

[yendacoder]

Hi,

I went quite far with this PR. Let me know if you feel like this deviates too much from your original vision, I'll publish my fork under a different name. I wanted to address two issues before using this library in my project: performance (scroll lag) and compatibility (icons support, specifically on Android: some icons while technically exist in system fonts, don't have a colored version)

So, here is the list of my changes (at least noteworthy ones):

Performace-related

• removed shrinkWrap property from the GridView (not sure what was the reason for adding it, but it can impact performance)
• refactored emoji cells into using InkWell instead of buttons for ButtonMode.MATERIAL mode (very noticeable scrolling performance boost)
• added ButtonMode.NONE. The biggest scrolling performance hinder comes out of using buttons inside GridView. Clients may now opt for using a plain GestureDetector for cells which gives butter-smooth scrolling, sacrificing visual touch response on cells.
• refactored TriangleShape into Decoration. Does not improve performance that much, but still saves a couple of widgets and makes the code cleaner.

Code simplification

• refactored raw data from lists of Strings to direct initialization of list of categories with Emoji objects (also applied the fix for the incorrect koala face position). This change allows to (a) maintain raw data easier, because now the emoji character, its name, and skin support are all in one place; (b) build a manageable way of providing custom set of emojis by the host app; (c) remove a lot of initialization code.
• refactored Android "support check", mainly to make it compatible with the new raw data structure.
• removed caching of supported emojis in SharedPreferences. In my tests, checking for compatibility only takes around 100-120ms. Considering that initialization is not blocking widget tree building, with Offstage widget happens before user gets a chance to see any picker UI, and retrieving the first instance of SharedPreferences alone takes 3 times longer, I think this optimization was unnecessary. Removing it allows to get rid of emojis versioning and simplifies initialization code even further.

Compatibility

• added a way to customize emoji set. Host app can use this to only allow emojis that are guaranteed to be supported by their target platforms. Or filter out emojis that users don't like for cultural reasons. Or add more emojis without waiting for library update.
• added a way to set a custom emoji text style. Allows to use custom font for emojis either with GoogleFonts package or with a font bundled with the app.
• added EmojiTextEditingController class that can be used to apply the same custom emoji font to a TextField input.
• provided an example of using library method EmojiPickerUtils.setEmojiTextStyle to apply custom emoji font to a RichText widget.

General usability

• added a setting to disable Android emoji compatibility check. If the host app is targeting platforms that are guaranteed to have support of all emojis, use custom font or a limited set of emojis, the check is unnecessary and may save some initialization time.
• provided an example of inserting emojis with EditableTextState that properly updates scroll position of a TextField, see comment for _onEmojiSelected method in the example for more details.
• removed progress indicator and related setting. Initialization on my not that performant device never takes longer than 0.5s and happens offscreen, having the progress indicator with customizable color only adds confusion, in my opinion.

Preview of the result.

[Fintasys]

@yendacoder wow, thank you for your huge contribution. I would have preferred each feature in its own PR for making it easier to review and judge if it should become part of the package. I know this can be difficult sometimes but it makes the live of maintainer so much easier and speed up things. Nevertheless I will try to review your PR, check your changes and will let you know about it. Also it probably conflicts with changes I had in my pipeline. Please give me some time to check it 🙇

[yendacoder]

Hi @Fintasys Yes, it should have been at least two PR's (some changes are quite interconnected, but, of course, not all of them), but I've made a lot of changes rather spontaneously. Sorry for that. I saw that your issues on the web branch (if it is the one you were referring to) are mostly related to custom fonts, which I think are working well with these changes. So, if we can get through review with this, updating the library for web support shouldn't be that difficult.

[yendacoder]

Thanks for the thorough review. I'll address some of the issues tomorrow, you had some excellent catches. The CI issue was caused by that koala PR indeed. Should be good now. I think it's good to make a major version out of this because of progress color setting deprecation (that would be a breaking change). Otherwise, it should be backwards compatible, even if we bring back caching.

[yendacoder]

I think that we are almost there with this PR. Before you make a new version out of it, I'd like to make another one that possibly can be included in the new version as well. It concerns search feature. I think that search itself can be outside of picker itself (otherwise, there will be too many settings with texts and layouts), so I'm planning to make another example dedicated to search specifically. But to make it more manageable, I think the library should expose a widget for emoji cells that host can use to display search results.

Let me know what you think.

[Fintasys]

I also think that the PR is good to be merged. I will also work on improving the documentation, the examples, updating version etc. I'm also curious about the changes you want to do to the search function! I would appreciate if you might can line out quickly some details in an issue first before starting a PR.

It seems CI(analyse) is not happy with you current code changes :D After addressing this I will merge this PR into master 👍

[yendacoder]

The lint issues have been fixed.

I've also created 2 issues to outline the next changes I have in mind.

I think that this PR can be merged.