.MGX file format

[BugA_the_Great]

download: mgx_format.zip (106 KB)

You have there a .mht file (html single archive), and a .xls file (excel file, improved version of html table, mostly in Japanese. I translated a bit to English, use Google Translate for additional translating).

Also, I suggest you to check A2LView here (by biegleux), Lister plug-in for Total Commander, open source AoC recorded game analyzer (written in Delphi). You may find some more nice info in its code

You may also check the AoE Development Group. Feel free to join.

 

[bkroxx]

nice i'll try to do something

 

[surreal]

Thanks. Very useful. I only bookmarked baris page, lucky someone saved this information.

 

[rox_rook]

Hi BugA_the_Great,

I read in raw .mgx in binary using C++, and saw many weird characters. What kind of format is that? By the way, if I remember correctly, AOC is produced by Microsoft Inc, how does it have Japanese characters, don't you know? I'm new to data compression and decryption, so could you provide some further information? :D Thanks in advance.

 

[BugA_the_Great]

Hi rox_rook, I haven`t dug to much into MGX structure, so I guess there are people on this forum that could help you more ( pokazeny_vysavac aka biegleux, jose, or _PkZ_ if he has the time).

The right way for reading MGX format would be HEX, there are no weird characters in there (only 0-9, A-F). You need to know which position (set of characters) means what in the file itself, so you could translate it to number, text, or whatever - and for that purpose is the excel file I gave, explaining the MGX file structure.

The file is in Japanese language because the guy who did the research on MGX file structure, Bari if I remember right, is from Japan

To see how others handle the compression and data parsing, check this, this and this topic (source code included).

 

[yech]

Very nice analysis! I wonder whether the network package during playing would follow the same format, anyone know anything about that?

 

[RoR_xFictionx]

Hey all!

I've recently discovered some research on the mgx file format i did back in 2006 on my old pc. I thought i'd share it with the community to try completing it together. I think alot of guys out here like cysion, scripter, mandrake88, ... may know alot of additional things, so we may be able to bring together the knowledge in a single document.

I hosted the stuff on https://github.com/stefan-kolb/aoc-mgx-format.

Would be cool if someone in touch with these guys may give them a nudge if they are interested to share their knowledge. Maybe this way we'll also see some new cool tools for aoc!

Cheers

Iam_Fiction

 

[SalzZ_dfs]

I've just remember about this file.. It has got "Achievement" block inside, but I've heard what this information become available only in UP.
Also, I see what aoczone have this information for UP records now.. Can you help me to get it?

I've stacked at line 386:
char unknown26[?] 不明 Unparsed data.

It's just before Achievements.

I don't know how much bytes I should skip after line 385 to get achievements data.

I hosted the stuff on https://github.com/stefan-kolb/aoc-mgx-format.

Seems like it's possible to get much more info from record, than we have now.. Here is a link to sc2 records anylyser: http://ggtracker.com/matches/5437719 (move your mouse over timeline)

 

[Idle Beaver]

I've just remember about this file.. It has got "Achievement" block inside, but I've heard what this information become available only in UP.

I don't know, what the Achievement block in this format description is, because it doesn't contain the achievements :P

UserPatch adds the post-game data structure to the rec file body. It looks like this.
UP puts it in as a 0xFF operation, see Biegleux's RecAnalyst for details: http://sourceforge.net/p/aoc-mgx-utils/ ... .pas#l1750

 

[SalzZ_dfs]

Hm.. thanks, I'd have a look.

Btw, I get to achievements data as

scenario_header_pos - 1817*num_players

The data here is really wrong. Gonna dig in the end :D

 

[SalzZ_dfs]

Seems like I get it. Thanks for help, Idle Beaver

 

[SalzZ_dfs]

One more question:

There is a command "0x77" (training unit).

Seems like no info about player who train this unit. But we can detect it with "building_id".

The thing is there is no info about "building_id" in "create building" block (0x66).

Also there is no block "0x66" in this ".MGX file format", but there is 0x166. Btw, the info I get with 0x166 command is not look like building creation 11

So no way to detect witch user create unit? Seems like "building_id" in "0x77" command is fine. Just need to get this id in "building creation" block.

 

[Idle Beaver]

Nope, the building IDs are not stored in the recorded game, but determined at play time. I don't think it's possible to find them without simulating the entire game.

(0x166 is just 0x01 followed by 0x66, 0x01 being Command and 0x66 being Create Building, which takes a player id and a building type id, both 16 bit ints)

 

[SalzZ_dfs]

ah ok
it's sad really :S

 

[RoR_xFictionx]

I'm not sure if the building ID is included there, if not, the behaviour is just incrementing the id for every unit and building during the game (https://github.com/stefan-kolb/aoc-mgx- ... 0Infos.txt). The description looks weird but maybe you can debug it with that (https://github.com/stefan-kolb/aoc-mgx- ... y/Actions/ - see 0x66). Sorry, but I can't remember more it's been too long ago...

 

[SalzZ_dfs]

Hm... interesting. Maybe I'd back to the project later.

For a now, I just get buildings timeline: http://rusaoc.ru/clientgames?gid=39343 (click on 4th stick "Здания")