 | |  |  |  |  |
| | | | | |
A MemCom database consists of a UNIX file system directory containing several files:
header The header file.
index[.old] The binary index file. Any
extension designates a previous version.
data.x The binary data file(s).
The MemCom database header file is a text file containing the top-level information. In the current version it contains the following parameters:
MEM-COM x.y z endian_flag address_size creation_date access_date header_size index_size data_size naddr taddr dstsz nentry nrsize cvties liq clt hmode set_name_size ndim_max nfdes descrl redncy
The table below describes all relevant parameters, some of them being either unused or obsolete.
| x y z | Major, minor, and patch level numbers. |
| endian_flag | Little/big endian flag: 1 means data are stored in little-endian format. 0 means that data are store i big-endian format. |
| address_size | Database creator address size (4 or 8). Not used anymore, since MemCom stores address in 8 byte format. |
| creation_date | Database creation time stamp. |
| access_date | Database last access time stamp. |
| header_size | Size of header in bytes. Not used anymore and set to 0. |
| index_size | Number of rows of index. |
| data_size | Size of data file in bytes + 1.
data_size designates the byte address where
data can be appended. Note that data_size
is not necessarily equal to the data file size. |
| naddr | Total size of data section (in bytes). |
| taddr | Not used. |
| dstsz | Size of index table structure (in bytes). |
| nentry | Number of entries in index table. |
| nrsize | Number of entries in index table that triggers index table resize. |
| cvties | Total size of cavities, i.e empty space, in data section (in bytes). |
| liq | Last entry in index table queried. |
| clt | Current index table length. Not used. |
| hmode | Index table search method. Currently not used. |
| set_name_size | Maximum set name size (in bytes). |
| ndim_max | Maximum number of array dimensions of any set. |
| nfdes | Not used. |
| descrl | Default dataset descriptor size (in bytes). |
| redncy | Not used. |
The file index contains for each
dataset its attributes (name, type, size, address, etc.). These
attributes are stored in IEEE little-endian
format, and their layout corresponds to the following (packed)
C-structure:
/*
* Dataset structure as read from and written to index file.
*/
struct dbsetsdb {
char sname[SET_NAME_LENGTH+1]; /* Name of dataset */
char type[4]; /* Type of dataset (E, I, D, etc.) */
char desclgt[4]; /* Descriptor size (bytes) */
char dspos[8]; /* Byte address of descriptor */
char spos[8]; /* Byte address of data part */
char ndim[1]; /* Number of dimensions */
char dim[MAX_NDIM][8]; /* Size of each dimension */
char status[1]; /* Status (0 inactive, 1 active) */
char hidden[1]; /* Hidden subset (0 no, 1 yes) */
char unit[1]; /* File unit (not used yet) */
};The length of the index (the number of entries) is stored in the header file.