Use machine-readable version number to determine current server's version

[smurthys]

This PR makes the following changes:

• Add a new function ClassDB.intServerVersionToString to convert a machine-readable pg server version number to string
• Use the server setting server_version_num instead of 'server_version' in function ClassDB.getServerVersion
• Remove support for distro suffix in function ClassDB.compareServerVersion
• Mark all functions in the script addServerVersionComparersCore.sql as IMMUTABLE
• Add test for new functionality
• Remove test for version numbers with distro-suffix

Fixes #263 #267

[smurthys]

I chose to convert current server's version from int to string for two reasons (instead of converting the input string version number to int):

• Only one function needs to change
• Considerably more logic is needed to ignore parts during comparison if int versions are to be compared. (need to bit-mask parts not of interest). This is compounded by the change in numbering starting pg10.

[KevinKelly25]

Thank you @smurthys for adding this useful function. Everything works well on 9.4 and 10.

The only small thing that I noticed was the use of v as a variable. While it is rather easy to guess the meaning of v, there aren't many places, if any, in ClassDB where it isn't immediately evident what a variable means. With v it took me a few seconds to make sure I knew the meaning. To me it doesn't fit into the nomenclature of ClassDB where it is normally easily readable. It seems like using version instead of v would fit better. However, it might be just me overthinking this and it is worth a small discussion.

[afig]

Looks great, thanks for updating the implementation @smurthys. The functions work as expected and tests pass on versions 9.6 and 10, and converting from int to string is certainly ideal all things considered.

I have two minor suggestions:

• Add a test in testServerVersionComparers.sql that uses the old versioning scheme but with a 0 second part (for example, asserting that ClassDB.intServerVersionToString(90005) = '9.0.5'). It seems like such a test will pass without issue, but the case is distinct enough that it might warrant a test.

• The comment on L68 of addServerVersionComparersCore.sql seems unfinished (ends with 'and` and does not point to a second source of information on the next line).

I also agree with @KevinKelly25's suggestion about replacing v with a more descriptive variable name.

[smurthys]

Thank you for the review. It seems like I forgot to replace the variable name I had used in the version I had developed in pgAdmin. Thanks for catching that.

I just pushed changes that address the three issues pointed out. Because a parameter name changed, please run the following query before running the revised script. This step is not required in production:

DROP FUNCTION ClassDB.intServerVersionToString(INT);

[smurthys]

Thanks for the reviews.