Tamas Balogh @itsbata

Android Developer @jaumo, previously @toptal, @autsoftltd. O{pi}nions are my own 🧅 Hungary Joined May 2010

Tweets

262
Followers

186
Following

238
Likes

853

Tamas Balogh @itsbata

4 years ago

@Zhuinden @ppvi My team has only 4 Android devs, not sure if this counts as "at scale". As for the problems, we have issues like missing shadows (no hardware rendering) and emulator hosting. Previously we had flakes for snackbars, cursors, and scrollbars but these had been worked around.

0 0 1 0 0

View Details

Tamas Balogh @itsbata

4 years ago

@Zhuinden @ppvi 124 test classes, 492 test methods, 896 screenshots Git (no LFS yet, but considering) Facebook Screenshot Testing tool verify task, also GH diff tool for review and ImageMagick's convert if I need to "debug" it I think all of them use Pillow except GH Nothing blocking

2 0 3 0 1

View Details

Tamas Balogh @itsbata

4 years ago

@VasiliyZukanov Such things happen, when infa engs consider forward fix is cheaper: engineering.fb.com/2021/10/05/net…

0 0 0 0 0

View Details

Tamas Balogh @itsbata

4 years ago

@arkann1985 @Zhuinden Why? Child presentation layer can say that button x is clicked, and then domain/application layer will decide what to do when it happens. If it decides that we should navigate, domain will simply build a url, and call an interface that hides navigation.

2 0 1 0 0

View Details

Tamas Balogh @itsbata

4 years ago

@arkann1985 @Zhuinden I don't see how does using deep links to address a destination in your navigation graph makes it child's concern.

1 0 0 0 0

View Details

Tamas Balogh @itsbata

4 years ago

@donnfelker @VasiliyZukanov That is exactly what I wanted to reply 👊 ppl will always have different opinions, and will always make mistakes. Some admit them, and some fool even themself. The former are the ones who have an impact.

0 0 4 0 0

View Details

Tamas Balogh @itsbata

4 years ago

@PeterEnyedi @zsmb13 Done

0 0 0 0 0

View Details

Tamas Balogh @itsbata

4 years ago

@flo_walther @VasiliyZukanov @p_chmielowski @plcoding Indeed, but one can create a "binding method", that listens on StateFlow<String?>,and shows/updates Snackbar when it changes, dismiss when null received and register a listener to set it to null when dismissed.

0 0 1 0 0

View Details

Tamas Balogh @itsbata

4 years ago

@flo_walther @VasiliyZukanov @p_chmielowski @plcoding You can simply have a StateFlow<String?>, set the value you would like to show, and when Snakbar is dismissed, set it to null. You will have one simple source of truth, which is consistent with what is rendered. Makes more sense than Channels

1 0 0 0 0

View Details

Tamas Balogh @itsbata

4 years ago

@soroush_l01 @Zhuinden @manuelvicnt @gamesdev_Dan @plcoding @p_chmielowski ViewModelScope uses exactly that. Just create the channel there.

0 0 1 0 0

View Details

Tamas Balogh @itsbata

4 years ago

@VasiliyZukanov I found the most reliable representation of a timestamp is the combo of an ISO8601 timestamp with offset plus the IANA TZDB id of the timezone. That one identifies an exact timestamp at an exact location.

0 0 0 0 0

View Details

Tamas Balogh @itsbata

4 years ago

@VasiliyZukanov I had burnt myself by using UTC timestamps. It only works when you don't need to do any calculations. But many times there are calculations that are not obvious like adding days and months to a timestamp for pagination or other rendering related tasks.

1 0 0 0 0

View Details

Tamas Balogh @itsbata

4 years ago

@Zhuinden Another thing to mention that ideally these classes you wrote are not in the same module, as one does not want their business logic to depend on a database engine, so this structure facilitates dependency inversion as well.

0 0 0 0 0

View Details

Tamas Balogh @itsbata

4 years ago

@Zhuinden But because it forces the presentation layer to contain business logic, that the use case should have internally.

1 0 0 0 0

View Details

Tamas Balogh @itsbata

4 years ago

@tuanchauict @CatalinGhita4 @Zhuinden Well in the previous example I would say 'I favor this approach because it facilitates better testability'. But if something is really readable then just say readable. E.g accessing a simple variable instead of observing something with a callback for one update is more readable.

1 0 1 0 0

View Details

Tamas Balogh @itsbata

4 years ago

@CatalinGhita4 @Zhuinden Or when there are two approaches where one is more flexible but the other one is easier to test then which one is "cleaner" in that sense? If someone value one trait more than the other would consider that one "cleaner". I guess we use it as a synonym of good

0 0 1 0 0

View Details

Tamas Balogh @itsbata

4 years ago

@CatalinGhita4 @Zhuinden I guess what you mean is Clean Architecture, which is well defined architectural guideline by Uncle Bob. Buy saying a piece of code is 'clean' I agree is meaningless. For instance the trait you had pointed out is independence.