Record executed plans on PostgreSQL

Fork

Fig regression test.

expected/convert.out had a wrong result based on a bug. Commit
9d7ea07912 fixed the bug with the test script being left alone. This
doesn't change the behavior so only source package with the next
release number is released.

Fix SPEC file for pg_store_plans11

The llvm subpackage wrongly rewires pg_hint_plan11. Fxied it. There'
no other differences so the dependency is described omitting the
release number.

Add SPEC file for PostgresSQL 11

The previous commit forgot to add the SPEC file for PostgreSQL11 and
some other files. Also Modify Makefile so that RPM files are sainly
generate.

Version bumped to 1.3

Changed version to 1.3. Only SPEC files for 9.6 and later are fixed to
generate 1.3 package since the last fixed bug doesn't affect the
eralier versions.

Fix boolean setter declaration

BOOL_SETTER ignores given name and always sets value to the member
"parallel_aware". When inner_uniue comes after parallel_aware, the
former overrides the latter and the latter doesn't get the right
value. As the result the following JSON plan is wrongly converted as
"Parallel Join".

select pg_store_plans_textplan('{"Plans":[{"Node Type":"Join","Parallel Aware":false,"Inner Unique":true}]}');

Currently only "Inner Unique" hits this.

Fix misspelling of pg_store_plans

In many places in the files pg_store_plans is misspelled as
pg_store_plan. Fix all of them.

Fix a typo of a message.

Fix a long-standing typo in an error message.

Follow a change in core.

Parameter of scanner_init has changed in core. Follow it.

Add the LICENSE file and update copyright notice of all files

The LICENSE file is added to explicitly declare copyright notice of
this software. Updated copyright notice on each files along with.

Support PG11

Support 64-bit queryid widen as of PG11.

Ignore trailing semicolon in hashing a query

Since PG10 pg_stat_statements doesn't store trailing semicolon in the
column "query". Normalization is basically useless in the version but
still usefull to match utility commands so follow the behavior change.

Bump up version to 1.2.

Supports PostgreSQL 10.

Add test for storing plans

There was no regtest for actual storing behavior. This commit add
that.

Modify the script to generate conversion test file

The script contained a query setting a multievent trigger using
transition tables. Split it into two triggers since such operation has
been inhibited.

Replace test files

Update test files so that all supported field names are tested.

Add support for new node types

Pg10 has some new nodes. This commit add support for them.

Support PostgreSQL 10

Followed the change of API of some executor functions.

Avoid duplicate warning on nonstandard escaping of strings

pg_store_plans emits duplicate warning with core parser for each
nonstandard string escape in expressions of a query. This commit stops
that.

Fix wrong indentaion of short json plan

pgsp_json_shorten insertes unnecessary line breaks for keys of
grouping sets. This commit fixes it.

Fix a crash bug caused by unkown aggregation type.

Unknown aggregation type caused a crash. This happens only on
development but should be treated as the same as other similar cases.

Print unknown properties in text representation

Foreign data wrappers may put explain properties unknown to the core,
such as "Remote SQL". This patch prints such properties in text
explains.

Support parallel query

Support Gather node and related properties.

Edit copyright notice

Edit year range of copyright notices.

Add RPM SPEC file for 9.6

To build a RPM file, do the following thing.

(Edit $PATH to point right pg_config)
$ make USE_PGXS=1 rpm96

Support PostgreSQL 9.6

Use user defined lwlock tranche from 9.6 instead of the old way of 9.5
and before. Follow the change of the third parameter of ExecutorRun.

Prepare RPM build environmnet for Version 1.1

Stored data is backward compatible with 1.0.

Fixed "Original Hash Backets", IO Timings, Sort Keys in text plan
representation.

Fixed text plan representation of "ON CONFLICT", "GROUPING SETS",
"TABLESAMPLE" and multi-targetted ModifyTables, which just were not
shown or looked broken.

Replace test scripts

Since some features added, the scripts and result files are repalced
with ones according to the current behavior.

Add support for TableSample in text representaion.

Text representaion of table sample properties to be shown properly
with this fix.

Sample Scan (cost=0.00..2768.00 rows=100000 width=4)
+ Sampling: system ('10'::real) REPEATABLE ('0'::double precision)

Allow handled properties not to have a short name

Shorten a json representaion allows the properties that the parser
doesn't knonw to be left unshortened. This is applicable to other
properties that the parser knows *but* don't shorten
intentionally. This is convenient some properties to be handled but
rarely seen.

Support ORDER BY and GROUPING SETS in text representation.

Text plan representaion omitted these items. This patch make them
properly shown.
Addition to that, GroupKeys (ON CONFLICT) introduced another way of
indentation of brackets. Make the bracket indentaion code behave
differently for the case.