The CBM DOS can write “relative data” onto disk, based on individual records. The advantage is that all space is allocated once and then randomly accessed if and when it’s needed – much like the tracks on a CD.
This approach is different from sequential files, which have to read or write the whole file in one go (Apple call this “reading and writing atomically”). So finding the 50th record in a sequential file can take some time, whereas retrieving record 50 from a relative file is just as fast as retrieving record 1 or record 157.
Let’s see how we can deal with such relative files in CBM BASIC 7.0. This works with all physical disk drives (1541, 1571, etc) as well as the SD2IEC card reader – but not with your trusty old Datasette (for obvious reasons).
This was revolutionary back in 1982!
Creating Relative Files</h3> Here we create a new file with a REL extension using the DOPEN# keyword. When we do this for the first time, we need to tell the operating system how much data we'd like to allocate per record. Anything up to 255 bytes is fine. We'll use 100 bytes in this example (the letter L followed by the length). <pre> 10 dopen#1,"rel test",l100 20 for i=1 to 100 30 record#1,i 40 a$="record"+str$(i) 45 print "writing record ";i 50 print#1,a$ 60 next 70 close 1 80 print "all done!"
When we create a new file, CBM DOS doesn’t know how many records we’ll need, but it will create as many as necessary on the file. Here we create 100 entries using a for loop and write them to disk using the PRINT# statement.
Notice that before saving a record, we must position the record pointer using RECORD#. That’s how our data is written to the correct position on disk. The whole loop will take a while to complete, but all space will be allocated on the disk as soon as we CLOSE the file.
To add more records, we’ll simply position the record pointer to a later entry – even if one does not yet exist. CBM DOS will create the last record and all records leading up to it so they can be used later, up to a maximum of 65535 (space permitting obviously).
Reading data from Relative Files
Much like with sequential data, we can either use INPUT# or GET# to grab the data for an individual record. INPUT# reads everything until a CHR$(13) is found and works much faster, while GET# reads data one character at a time and is a lot slower.
10 input"which record shall we retrieve";r$ 20 dopen#1,"rel test" 30 record#1,val(r$) 40 input#1,a$ 50 close 1 60 print:print a$:print 70 print"need another one? (y/n)" 80 getkey k$:if k$="y" then goto 10:else end
Here we ask the user which record we want to retrieve. Next we open our relative file using DOPEN#, position to the desired RECORD# and then read in its value using INPUT#. When we’re done we close the file and ask if the user wants to retrieve another record.
While this type of data access is quick and convenient, it doesn’t help much unless you know which record is stored at which position. Think of an address book application: to find “Johnny Appleseed” can’t be done unless you sift through every single record until you find him.
Commodore therefore suggest to use sequential files alongside relative files, in which things like a last name could be saved together with the record number. Likewise, another sequential file could hold all records for the first names, and appended accordingly when a new record is created, or replaced when updated.