Simcity 2000 Special Edition

• Return to the list of games

Choose archive extension:

IDX + DAT

• Format Type : Archive
  
• Endian Order : Little Endian
  

Format Specifications

// for each file


uint32 {4}   - ID

uint32 {4}   - File Offset




MultiEx BMS Script

Not written yet

Notes and Comments

• The *.idx file contains the directory, the *.dat file contains the file data

Compatible Programs

• Game Extractor
  

SC2

• Format Type: Interchange File Format (IFF)
• Endian Order: Little Endian

Format Specifications

// HEADER

char {4}	- Chunk Type ID ('FORM')
int32 {4}	- Total count of bytes in file minus first 8 in the header
char {4}	- File type (always 'SCDH' for SimCity 2000)

// Begin segments/nested chunks. For each nested chunk:

// Segment header

char {4}	- Segment type
int32 {4}	- Byte count of the data in this segment (excluding this header)

// Segment data, begin sub-segment bytes (let x = the byte's value)

byte {1}

// If 1 <= x <= 127:

byte {x} - Raw, uncompressed data

// If 129 <= x <= 255:

byte {1} - The following byte repeats x-127 times.

// Repeat for the rest of this segment's data length.

Notes and Comments

• This is the file format in which cities are stored. See the IFF page on Wikipedia (linked above) for more information on the format, although it is essentially a generic container file format.
• All of this information is thanks to the research done by David Moews, with a document here.
• After the header, the rest of the file is made up of chunks containing an 8 byte header followed by its data. Almost all segment data is compressed using a run-length algorithm. Specifically:
  • The compressed data is a series of two kinds of chunks
  • In the first kind, the first byte equates to an integer 1 to 127. This means the byte is counting how many data bytes follow.
  • In the second kind, the first byte equates to an integer 129 to 255. If you subtract 127 from it, you end up with a count of how many times the following byte is repeated.
  • Chunks with a first byte of 0 or 128 never seem to occur.
  • Therefor, the maximum size of an uncompressed sub-chunk can be 127 bytes.

Segment Types

• The following table is a list of segment types in the order they typically appear. Unless noted, they are compressed using the above algorithm. The lengths given are their uncompressed lengths.

Seg. type	Length (bytes)
CNAM	32 (Uncompressed)
MISC	4800
ALTM	32768 (Uncompressed)
XTER	16384
XBLD	16384
XZON	16384
XUND	16384
XTXT	16384
XLAB	6400
XMIC	1200
XTHG	480
XBIT	16384
XTRF	4096
XPLT	4096
XVAL	4096
XCRM	4096
XPLC	1024
XFIR	1024
XPOP	1024
XROG	1024
XGRP	3328

Additional notes on segment types:

ID	Comments
CNAM	The name of the city, uncompressed, and also optional it seems. When it's present, the length byte is a number 0 to 31, with that many bytes of city name. It's padded to 32 bytes with zeroes.
ALTM	Altitude map, which is uncompressed. Each "square" is two bytes, with them being scanned in top to bottom, and from right to left in each row. In my testing, this seems to mean that the first square is at the northern-most corner of the screen. Following that, squares proceed to the south-west, or left in this case. Let every two bytes represent a 16 bit integer, MSB first, such that bits 4-0 represent the altitude from 50 to 3150 feet. Bits 6-5 are unknown. Bit 7 seems to be set if a square is covered by water. Bits 15-8 are unknown. Notes from my own testing (in SCURK 1.0): By default, it seems that the default altitude of a square (starting in SCURK) is 4 "units" high, as is represented by a binary 00100. Raise that square by one, and it will become decimal 5, or binary 00101. At some point along a row of empty level squares, hex 0x0084 squares show up in groups of 22 bytes, or 11 squares. This is bin 10000100, which is still the usual four squares high, although bit 7 is set despite the square not being under water. Furthermore, in a totally unmodified city (with only the name/year/budget having been set), the first eight bytes of ALTM data (or four squares) have this hex 0x0084 pattern, followed by 20 bytes (10 squares) with the hex 0x0004 pattern. Eight more bytes follow that with hex 0x0084, and the alternation continues although the number of squares in each set does not seem to correlate with anything. Upon generating another city with a different name, but with all other parameters the same, the alternating pattern is also exactly the same. Starting with hex 84 then hex 4 and repeating, the number of squares is: 4, 10, 4, 13, 3, 3, 31, 25, 8, 27, 3, 11, ... Bit 7 apparently does not seem to indicate water coverage, although it could indicate that water was on this tile at one time. Raising terrain as in the previous example does appear to disturb this pattern, with squares of hex 4 becoming more abundant. Adding a pond of water to a square sets bits 7 and 6. Interestingly, bulldozing the pond, or changing the elevation of the square (which visually removes the pond) does not unset those bits, despite the pond apparently being gone. Further testing conducted in SC2000 Win 95 version: Lowering the tile but adding the pond back leaves only bit 7 set, unsetting bits 6 and 5 if they were set. Raising it once again and adding the pond sets bit 5. Ultimately, it seems clear that bits 7 thru 5 has something to do with water. What exactly that is remains unclear; perhaps it was supposed to be used in a feature that was never implemented.