Add has_view(:schema, :name) and hasnt_view(:schema, :name)

[Godwottery]

This is PoC for #219

I haven't been able to test it due to lack of suitable machine.

[Godwottery]

I got stuck on the unit-tests. I can not seem to get them to line up.
Could someone please point me in the direction where I can find the cause.

test/sql/hastap.sql .. 1/896 
# Failed test 115: "has_view(sch, view) should pass"
#         have: false
#         want: true
# Failed test 116: "has_view(sch, view) should have the proper description"
#         have: tables
#         want: View "information_schema.tables" should exist
# Failed test 119: "has_view(sch, non-existent view, desc) should have the proper description"
#         have: __SDFSDFD__
#         want: View "foo.__SDFSDFD__" should not exist
# Failed test 136: "hasnt_view(sch, view) should fail"
#         have: true
#         want: false
# Failed test 137: "hasnt_view(sch, view) should have the proper description"
#         have: tables
#         want: desc
# Failed test 140: "hasnt_view(sch, non-existent view) should have the proper description"
#         have: __SDFSDFD__
#         want: View "foo.__SDFSDFD__" should not exist
# Looks like you planned 896 tests but ran 893

[nasbyj]

For one, you subtly changed the test; it used to explicitly test the description field in both cases, now it doesn't.

More importantly, you're changing the calling behavior. Currently, we have:

~@88e9fe532b8f.ant/87749# \df has_view
                         List of functions
 Schema |   Name   | Result data type | Argument data types | Type
--------+----------+------------------+---------------------+------
 public | has_view | text             | name                | func
 public | has_view | text             | name, name, text    | func
 public | has_view | text             | name, text          | func
(3 rows)

Which ensures that the 3 different functions can be uniquely distinguished by the database just by looking at the number of arguments. Your patch is adding a (name, name) function, which can not be readily distinguished from the (name, text) version. A quick test on 11 looks like the engine will default to (name, text), preserving existing behavior, but we need to have explicit tests for that.

[Godwottery]

As far as I could tell, has_table has the same problem type. There are (name, text) and (name, name).
I think that the tests indicate that the text one is the default one. The name version is tested with a '::name' typecast.

[Godwottery]

Using the ::name typecast works for old versions of PostgreSQL (9.1 and 9.2)
This does not work for newer versions.

I think this confirms findings of @nasbyj. The (name, text) seems to be prefered.
I would like to argue that there is explicit tests for this -
has_view( 'pg_tables', 'yowza') should return 'yowza' if (name, text) is run. That is explicitly tested for.

Are there any ideas on why the explicitly typecast versions are not begin run correctly?

text seems to be the preferred datatype. Use select typname, typispreferred from pg_type where typcategory = 'S'; to check this. I could not find any 9.1 installation, but checked on 11 and 8.4 where text is preferred.

[nasbyj]

I think you're mis-reading the test output (which is super-easy to do). The clue is at the very end:

These test targets failed: 'clean updatecheck'

If you search for "updatecheck" in the test output you'll find the section where that actually runs, and that's where the failures are, which actually makes perfect sense because you haven't added an update script for doing ALTER EXTENSION pgtap UPDATE.

This is a bit of a failing in our release process... after we tag a release we should really increment the version number and create the relevant pgtap--x.x.x--y.y.y.sql file in sql/. I'll go do that in master tomorrow. If you merge that change and then add the new functions to sql/pgtap--1.1.0--1.2.0.sql then tests should start passing.

(The reason 9.1 and 9.2 tests are passing is because ALTER EXTENSION UPDATE didn't exist back then, so we don't test updates there.)

[nasbyj]

I've committed the necessary changes to master; please merge. You'll get a conflict on the update script; please keep the definition for pgtap_version() at the top of the file.

[Godwottery]

That fixed the tests. Thanks for the help.