Localization scraping is incredibly broken

[cloakedyoshi]

Issue Type: Other
Impact: Very High
Frequency: Often
Date First Experienced: 2021-04-10 01:04:00 (-04:00)

Reproduction Steps:

1. Have a game with various strings
2. Enable “Automatic Text Capture” and “Use Translated Content”
3. Observe strings are often scraped in really odd places - as in not the entire string

Expected Behavior:
The entire TextLabel’s text string is scraped, not a specific part of it.

Actual Behavior:
Occasionally specific parts of a TextLabel are scraped - not the entire string. I cannot tell if this is due to the new MaxVisibleGraphemes property, RichText, or something else. It seems to be a somewhat recent regression. This makes manual localization incredibly difficult and cluttered.

English/original version. Each one of those strings in that list is a single TextLabel.

Neat, it’s autolocalized in Portuguese… but wait… the person doing manual localization cannot find the string!

So let’s take a look. I added the top two strings manually, everything underneath is what was done automatically. The entire string is not scraped for some reason!

This is my game that has the most issues with this:
https://www.roblox.com/games/243079752/The-Day-the-Noobs-Took-Over-Roblox-2

[0xabcdef1234]

Hi, we are aware of this issue and would to double check some of your usage on those text features:

• Are you using RichText for those strings?
• Are you applying any text animations for those strings while displaying them with string.sub()?

We have some backend processing for localization services too as for text capture parameter content, that means if you have Is your {something} within your text entry or you have mass Is you {custom_items}, it will then recognize your partial string as parameters.

We will take a look at this issue asap.

[cloakedyoshi]

@0xabcdef1234

• Are you using RichText for those strings?

Nope.

@0xabcdef1234

• Are you applying any text animations for those strings while displaying them with string.sub()?

Nope. These strings are in unanimated UI elements. No old-school or modern typewriting effect involved.

@0xabcdef1234

We have some backend processing for localization services too as for text capture parameter content, that means if you have Is your {something} within your text entry or you have mass Is you {custom_items} , it will then recognize your partial string as parameters.

Thanks for the context! These are not generated via a script if it helps. They’re separate pre-written TextLabels. No coding involved whatsoever on my end.

[0xabcdef1234]

yeah we could have some logic during generating table that might affect this. Do you mind help checking is there any entry contains Is your in your auto-scraped table?

[cloakedyoshi]

I do not see any form of “Is your” on its own in the localization tables. It appears to be what comes after it.

For safe measure I also checked the missing part of other strings (“Are” for example) and nothing comes up.

Searched “is your” - I added these manually, so there’s nothing scraped.

Searched “bug report” - a mix of manual and automatic strings.

[YuckyBucket]

Hey @cloakedyoshi. It looks like your issues are caused by some weird parameter matching. Sources like “{1}” or some major matching source like “A{1}” can cause the problems you are seeing. We recently improved some of the automatic parameterization. Removing those bad sources and allowing things to rescrape should help resolve your problem.

[cloakedyoshi]

Thanks! For context what is the proper formatting of parameters? I have had this table around for a while and as a result it has a lot of old manmade and automatically-generated parameters.

[YuckyBucket]

Some info on parameters can be found here: Localization | Roblox Creator Documentation

[cloakedyoshi]

I have created a feature request in regards to this issue, as it resulted in many complications:

Automatic Text Capture should automatically be disabled in case of parameter mismatch

Not sure what category this belongs in, going with Engine Features but feel free to move it. I recently ran into an issue that gunked up my localization table [image] Notice the partial strings of the exact same content. The Cause The issue was caused by weird parameter matching. Aftermath This parameter issue was caused by having an old localization table that had gone through many iterations of the tooling. To my knowledge it was not user error (but that’s neither here nor there). I h…

A quick TL;DR:

• This issue added ~1300 junk/corrupted entries through the Automatic Text Capture tool
• It took me over six hours to remove those entries, and I probably didn’t even get them all