Misleading Filesize Note?

[Jaina Proudless]

I have noticed a while ago that a cool model has a much larger filesize than it should have.

Below the download screen it says it is only 99.59 kb, quite a kick ass filesize indeed. The compressed file is the same size but the actual model is a lot larger after it is actualy extracted, somewhere around 300 kb (a bit too tired to specify exactly right now).

In the comments for the model it is wronfully commended for the low filesize and one user also notices the real filesize and points it out.

I only found this on this model for now. I only noticed it because my map turned out a lot larger than I calculated. I always look at the filesize that is noted below the download button as a decisive factor when download models and if the file inside the unpacked file is bigger it makes the shown size redundant. Furthermore, the ratings the users gave and the impressions the model left are slightly undeserved due to this... just a note... but I care mostly just for the utility of the shown filesize as I rely on it.

I hope that this is an exception and most models show the right filesize. If not this makes the shown filesize not very useful at all.

 

[Ralle]

I just tried reuploading. The date was fixed. I guess an old upload script looked at the .zip file.

 

[Dr Super Good]

20:33, 30th Oct 2007

If you would have posted this on WC3Search 8 years ago before Hive merger I am sure the author would have really appreciated it.

I only found this on this model for now. I only noticed it because my map turned out a lot larger than I calculated. I always look at the filesize that is noted below the download button as a decisive factor when download models and if the file inside the unpacked file is bigger it makes the shown size redundant. Furthermore, the ratings the users gave and the impressions the model left are slightly undeserved due to this... just a note... but I care mostly just for the utility of the shown filesize as I rely on it.

The size is not useful anyway. Compressed size (with WC3 compression) would be more useful since that would give you the actual space the model will take up. Uncompressed size is meaningless as some models compress very well while others hardly compress at all. Textures are the same since some palleted BLP files appear large uncompressed but compressed they are smaller then JPEG compression for a similar quality.

 

[Jaina Proudless]

If you would have posted this on WC3Search 8 years ago before Hive merger I am sure the author would have really appreciated it.


The size is not useful anyway. Compressed size (with WC3 compression) would be more useful since that would give you the actual space the model will take up. Uncompressed size is meaningless as some models compress very well while others hardly compress at all. Textures are the same since some palleted BLP files appear large uncompressed but compressed they are smaller then JPEG compression for a similar quality.

Hmmm quite interesting... Do these imports compress when they imported or after map optimisation or both? Is it somehow possible for the hive or hive 2 to show the exact extent of the compression for each model?

Also aren't uncompressed sizes useful as they shows you the maximum size the model will take up allowing you to make calculations without fear of exceeding a desired limit?

 

[Dr Super Good]

Do these imports compress when they imported or after map optimisation or both?

They compress along with all files in the MPQ archive upon map save. Without such compression terrain on a 480*480 map would take up >4MB by itself.

Some map optimizers compress better than the default World Edit map save. This is because they either use a different compression algorithm (WC3 supports many) or they use a better compression implementation (different vendor, maybe more modern with better algorithms).

For example people report Google's compression libraries do a better job (better compression ratio) than the implementation World Edit uses which can save quite a number of kilobytes.

Is it somehow possible for the hive or hive 2 to show the exact extent of the compression for each model?

Technically it would be possible. All that is needed is to shove it through some form of compression library (such as from Google) and then store the size of the result as "compressed size". This would not be completely accurate (for reasons mentioned above, different libraries and vendors and MPQ settings will cause size to vary) but is a far closer estimate of actual compressed size than the uncompressed file size currently is.

After the map is put through an optimizer with same compression library the file sizes should be within a kilobyte of the compressed size. The reason for this is because the MPQ archive format has overhead for the file table and hashtable. Additionally unless put through an optimizer the result might be much larger than the compressed size due to block fragmentation (optimizers often set block size to something insane like 2MB to eliminate fragmentation allowing for better compression).

Also aren't uncompressed sizes useful as they shows you the maximum size the model will take up allowing you to make calculations without fear of exceeding a desired limit?

To some extent they are, however under that model a 480*480 map would already have 6.5 MB used of the 8 MB limit for the terrain files alone. In reality such a map will more likely only have 2 MB odd used for the terrain, less if simple terrain is used.

 

[Ralle]

Sure, compressed size would be ideal and it's possible. I will add it to my list.