Allow the user to specify the baseline log file

udi-speedb commented 1 year ago

Currently the baseline log file is determined automatically by the parser, based on the version of the parsed log and the available baseline logs.

A user might be using a different baseline than what is found in the parser which they want to check against and in that case, its good to be able to supply it externally.

To implement this the following is needed:

Adding an optional parameter to the parser's command line arguments.
Verifying that the specified file exists and is a valid log file
A baseline should only contain options for the default cf. Handle files that are not compliant (ignore other cf-s? Reject? Other?)
Write unit tests to test this capability

udi-speedb commented 1 year ago

@Yuval-Ariel - Please update the description above and add motivation / use case for this feature request. Thanks

Yuval-Ariel commented 1 year ago

i've updated the description. The expected behavior regarding comparing a baseline LOG file which has multiple cfs is indeed problematic. i think the expected behavior should be to compare cfs which similar names if they exist and leave out the rest. e.g. baseline has default, cf_1 cf_2 and the LOG parsed has default and cf_1. so compare default and cf_1 and leave out cf_2 from the baseline diff.

udi-speedb commented 11 months ago

i've updated the description. The expected behavior regarding comparing a baseline LOG file which has multiple cfs is indeed problematic. i think the expected behavior should be to compare cfs which similar names if they exist and leave out the rest. e.g. baseline has default, cf_1 cf_2 and the LOG parsed has default and cf_1. so compare default and cf_1 and leave out cf_2 from the baseline diff.

@Yuval-Ariel - This will then work differently than how it currently works. At the moment, the cf-s of the parsed log are compared to the default cf in the baseline. Actually, there is an assumption that there aren't any other cf-s in the baseline (no explicit check).

For example, if we wanted to compare a parsed Speedb log to a RocksDB baseline log that we chose (assuming this is desireable), it will just compare the default cf-s and ignore all of the other cf-s in the parsed log.

What are the use cases that make this issue needed? When does it make sense to compare 2 logs that share some cf names but differ in others?

Yuval-Ariel commented 11 months ago

@udi-speedb , the way it currently works is also fine

speedb-io / log-parser

Allow the user to specify the baseline log file #12