Re: [open-cobol-list] Team setup

[henry . eon]

Hello,

I read this point in the TODO file :


** New backends for INDEXED files

 - Any alternative to BerkeleyDB?
 - Use DB (PostgreSQM, MySQL, etc.) as backends


Actually, I wrote my own implementation of indexed and relative files.
I used it for 3 years for personnal needs.
It could deal with big files (20 millions records, and more than 500 Mo) with
good performances. (based on trees and mmap)
It has some deficiencies and must be improved : keys are single fields, delete
is not ready, etc .. And surely it has some bugs.

Do you think it could be an alternative to BerkeleyDB. Is it a good idea or not
? My opinion is yes but I am partial. And there is still a lot of work to make
it a solid product. It would be a full job, and doing that I could not do
another thing. Furthermore the task we subscribe is COBOL compilation, not file
management.

Yet, I think that a COBOL compiler should have its own implementation for ISAM
and should not depend on third products like BerkeleyDB. (Althought I am a great
fan of it).
It has to be self-sufficient.
And DB (MySQL, Postgres, ...) induce constraints and minimum administration
tasks which I think are not compatible with the simple needs of an ISAM tool.

What is your opinion ?

Cheers
Henry