Re: [Bug-apl] What is the right way to layout data in a component filesystem

[Peter Teeson]

It depends on your requirements. 

The Sharp APL component file system was very often used as a one component per data class. 

There could be several files open depending on the application needs. 

This was before enclose/disclose. (Think SQL tables together making a DB).

The file system had a time stamp that could be used to highlight that a component had been modified. Think audit trail.

In fact in an amusing incident we had programmed the board game Diplomacy and had players from around the globe.

Everyone, except the player designated as the operator, had read and append only permission.

In one game the operator cheated and replaced his move after everyone else had made theirs.

We caught him because of the timestamp being out of sequence.

So it really depends on your needs. For example the granularity of the data that you might want to change, load a little or load a lot.

Today’s computers have so much RAM and HD/SSD coupled with modern apl that the best practices have probably changed.

respect

Peter

On Jul 26, 2016, at 8:57 PM, address@hidden wrote:

Hi bug-apl,

If I have three items, should I

1. Have three component files and use the library's append, delete, etc methods to update them. An append would make a new component in the component file.

2. One component file. Only use components 1, 2,  and 3. Then write methods to append, inserts, delete methods for each of these components?

3. Some other way entirely!

I haven't come across any papers about component file systems. I guess they are pretty APL specific.

-alex