Improvements for implicitly GENERATED fields

[midenok]

Currently implicit fields are named sys_trx_start, sys_trx_end. There should be ability to define arbitrary default names for them.

Column names should be mutated by prefixing _ to both sys fields in case of name clash with existing columns until there will be no clash.

In alter.test:
Test adding system versioning when a table already has sys_trx_start column, say, INT or VARCHAR.

[vuvova]

I don't think you need to bother implementing configurable default names. We have automatic generated names for indexes, foreign keys, check constraints — none of them are configurable. The user either specifies a name in CREATE TABLE or she doesn't, and then the server automatically picks up a non-conflicting name (e.g. with _1, _2, etc suffixes).

I'd suggest to be consistent here and follow existing behavior.

[midenok]

It may perfectly keep consistency with existing behaviour if f.ex. variable is not set. It would be good for me to have control of these names to spare some typing.

[vuvova]

if the whole purpose of this variable is to save you some typing, then you can add a variable only in debug mode. Or add it, and at the very end of the project remove it and search-replace tests to use longer name (I've did something like that a couple of times).

[vuvova]

Please, no "mutation". First, it does not belong to system versioning feature, these fields are simply created SYSTEM_INVISIBLE and they exhibit all behavior of SYSTEM_INVISIBLE fields.

Second, SYSTEM_INVISIBLE fields have well defined documented names (like ROWID), they are not renamed automatically (otherwise they'd be rather difficult to find and use).

[midenok]

Too many reserved names leave less freedom for application. This may be especially important for transparent versioning mode. DBMS is system that serves application needs, but not vice versa. But this is not critical and may be postponed until real request.