FWD

This task is about Non-Standard Save Cache Implementation .
Tijs will do the required changes, Eric needs to check/correct.

#8437-56: Greg Shah wrote:

Tijs Wickardt wrote:

I suggest to do the edits myself, and notify Eric to review the changes.

OK

todo:

Document requested change to Eric F at #3912-55

We changed it to:
save cache complete mscdata to save_cache_mscdata_dummy_output.txt.
which I predict will convert to:
SchemaCheck.run(new character("mscdata"));

Noticed by Ovidiu at #8434-26:

UDFs (generic functions used by FWD queries, also support for WORD indices)

Tijs, while I was in the doc for #8653, I reviewed your changes. I updated the UDFs section to point to the existing documentation about these. I also gave some examples of the recommended, "simple" installation approach. I left all of your other examples in place. I assumed they are correct and tested.

As to the SAVE CACHE syntax example, note that in the intro section of the doc, it mentions that the parameters after the logical database name are ignored. The feature was implemented to convert existing instances of SAVE CACHE statements where those last parameters would already be there, but they serve no purpose in FWD's implementation.

That being said, I haven't tested this feature in a while and I'm not set up to easily test it now; however, if one is adding SAVE CACHE statements only for FWD's benefit (as you are within the preprocessor conditional in your example), I think you can safely omit those additional parameters. This merits testing before you change the syntax example in the doc, though.

Thank you Eric, I added a small addition to the initial value section. Please review.

Eric, the wiki mentions "Such a scan is also done of all mutable databases at server startup" . Reading Ovidiu's release notes, I believe this is no longer true.

Tijs,
I do not think that is correct. The phrase from wiki is correct: a full scan is performed at server startup and the metadatas for mutable databases are updated with the tables extracted from SQL. I am sorry for the possible confusion. Which of my notes caused you to believe otherwise?

Ovidiu Maxiniuc wrote:

Tijs,
I do not think that is correct. The phrase from wiki is correct: a full scan is performed at server startup and the metadatas for mutable databases are updated with the tables extracted from SQL.

Thanks, all's good then.

I am sorry for the possible confusion. Which of my notes caused you to believe otherwise?

It was #7847-421 , where you mention:

• skip loading tables from mutable databases at server start-up, they will be loaded on request;

Tijs Wickardt wrote:

It was #7847-421 , where you mention:

• skip loading tables from mutable databases at server start-up, they will be loaded on request;

This refers to loading tables and fields into SchemaDictionary. However, to do so, the information must be readily available in the other meta-structures. And this information is performed at server startup or at request from SQL by SchemaCheck, as explained in documentation.

Ovidiu Maxiniuc wrote:

Tijs Wickardt wrote:

It was #7847-421 , where you mention:

• skip loading tables from mutable databases at server start-up, they will be loaded on request;

This refers to loading tables and fields into SchemaDictionary. However, to do so, the information must be readily available in the other meta-structures. And this information is performed at server startup or at request from SQL by SchemaCheck, as explained in documentation.

Clear, thank you.