Strip metadata from PNG mods/default textures

[Zweihorn]

Reduces significantly the size of PNG texture files found being sized 4 kB and above.
The selected files were processed with the optipng -o2 -strip all <file> command line.

The OptiPNG program shall attempt to optimize PNG files, i.e. reduce their size to a minimum, without losing semantic information. In addition, this program shall perform a suite of auxiliary functions like integrity checks, metadata recovery and pixmap-to-PNG conversion.

A brief test with a MT 5.0.1 client with digging dirt and emptying buckets of lava, river water and water showed no apparent issues but burned all trees nearby.

Hope this helps.

[Calinou]

If we're going to optimize all images (and increase Git history size), we might want to optimize it even further. Using oxipng, this can be done with oxipng -o6 --strip --zopfli **/*.png. (Or with parallel, parallel oxipng -o6 --strip --zopfli ::: **/*.png.)

[Zweihorn]

Oh sorry, the listing inside the commit was only ment for transparency.

1. Does this bother you or GitHub?
2. Is a reduction of texture files a pointless task?
   However, the file heart.png showed a 97.81% decrease. New size is 325 bytes from 15 kB before.

Naturally, there may be other tools. The optipng tool has other options too and I intentionally avoided to touch all files.

[Calinou]

Does this bother you or GitHub?

By "history size", I meant the total file size of the repository once you clone it. Unlike text files, Git doesn't handle binary files using delta compression; all revisions of a binary file will be downloaded when you clone a repository. A large repository takes more time to clone for people with slower connections; hence, we should try to optimize images once and do it well 🙂

Is a reduction of texture files a pointless task?

Definitely not, it matters when people join a server for the first time.

[sofar]

It's best done right before a major release :)

[Zweihorn]

we should try to optimize images once and do it well

It's best done right before a major release

So you propose to delay for an indefinite period?

You must be joking. No offence but I would call it rather unfortunate that this was missed for the 5.x major release and you want to wait for 6.x apparently.

Please note there are several more files to be reduced in size by 50% or more which are in other folders of the mods tree besides the 'default' mod.

[paramat]

we might want to optimize it even further.

Depends what this further optimisation includes.

Stripping metadata is well worth doing as it can cause a large file size reduction. However, apart from that, these textures have been optimised several times in the past and small additional optimisations are not important.

So this PR is essentially fine as it is, but it's better to wait until just before release of 5.1.0, which will only be a few months, and include any extra textures in the PR.
👎 For this PR.

I reopened the reminder issue #1831

[Zweihorn]

Thank you all for the consideration and the advice. I appreciate the 'maintenance' tag.
Easily understood and waiting for a minor release seems to be reasonable.

This PR and the list of files took already into account the majority of the PNG files can be called optimized. The optimization rates are 80% and better for all of the 20 files selected. So 'mods/default' was overlooked in the past release apparently.

IMHO one should not seek to optimize too far and should avoid tinkering too much with the optimization levels of the tools. AFAIK from some tests on the mods tree the outcome may be limited to the majority of files.

What about an optimization threshold? Taking into account the a.m. 'history size' argument I would propose to set a 10% threshold. Thus intentionally ignoring the huge majority of the PNG files and aim for a reasonable list of files which can be sufficiently optimized.

See you again just before release of 5.1.0 in a few months hopefully. Happy hacking.