Byte sizes need to be parsed with invariant culture

[zsoi]

On systems using a different locale than english, the default locale used to parse floats can produce incorrect numbers from a english-written build layout, e.g. 81.27kb may be parsed as 8127 kb with a german locale.

To prevent this from happening, machine-readable numbers should always be written and parsed with the InvariantCulture. This gives reproducible results regardless of the active locale on the thread.

This change just adds the InvariantLocale onto the float.Parse() invocations for sizes.

[pschraut]

Thank you for looking into issue #1 and the PR 👍

While the number parsing itself is fixed with your PR, the underlying problem remains unfortunately. Perhaps you have an idea how to fix it.

The problem is that Addressables stores numbers using a comma in a comma-separated list. This makes it unreliable to split the list-items from actual numbers, because the comma is used for different things.

Here is an example:
Archive test.bundle (Size: 1,79KB, Compression: Lz4HC, Asset Bundle Object Size: 280B)

The list-items or pairs are actually:

• Size: 1,79KB
• Compression: Lz4HC
• Asset Bundle Object Size: 280B

You can think of what's in-front of the colon as key (i.e. "Size") and what's on the other side as value (i.e. "1,79KB"). The code that parses this list can be found in BuildLayout.cs in the ReadAttributes method.

Since Size: 1,79KB contains a comma, the code handles the pairs more like this instead:

• Size
• 1
• 79KB
• Compression: Lz4HC
• Asset Bundle Object Size: 280B

Everything else in Unity is serializing its data using an English locale, but Addressables for some reason isn't. Maybe this can be avoided by adding some trickery to the ReadAttributes code, but I was running out of ideas.

[zsoi]

As mentioned in #1 this can be easily fixed with a custom fork of the addressables package. I filed a bug report with Unity and they may address this issue in a next release but from what I've heard they will scrap the textual build layout format for good in the next minor release and go with something more machine-readable by default.

[pschraut]

Oh sorry, I didn't see your comment in the issue before. Yes, makes sense. Thank you for the PR!

[gustavopsantos]

@pschraut Cant we use ', ' (comma + space) as a delimiter to parse the separated values? This way we would separate was is number and what is CSV delimiter.