[Dbix-class] Bad/missing joins on generated SQL (0.08112)

[Nilson Santos Figueiredo Jr.]

So, it seems like I'm always prone to this kind of stuff.
Well, anyway, today I moved some code from a server with DBIC 0.08107
to a new server, with the latest DBIC 0.08112 and some joins stopped
working.

Specifically, this query doesn't produce the desired results anymore:

  my $rs = $schema->resultset('FilingPDF')->search(
      { 'me.file_path' => undef, processing => '0', ticker => { '!=', undef } },
      {
          prefetch => { filing => 'company' },
          join     => { filing => { company => 'tickers' } },
          group_by => 'me.id'
      },
  );

Generated SQL on 0.08107 (and also on 0.08102):

<sql>
SELECT me.id, me.filing_id, me.file_path, me.queued_on,
me.processed_on, me.processing, filing.id, filing.company_id,
filing.title, filing.type, filing.date, filing.url,
filing.retrieval_date, filing.file_path, filing.processed_on,
filing.reference_date, filing.reference_quarter, company.id,
company.name, company.short_name, company.company_type,
company.market_type, company.last_update FROM filing_pdf me  JOIN
filing filing ON filing.id = me.filing_id  JOIN company company ON
company.id = filing.company_id LEFT JOIN company_ticker tickers ON
tickers.company_id = company.id WHERE ( ( me.file_path IS NULL AND
processing = ? AND ticker IS NOT NULL ) ) GROUP BY me.id
</sql>

Generated SQL on 0.08112:

<sql>
SELECT me.id, me.filing_id, me.file_path, me.queued_on,
me.processed_on, me.processing, filing.id, filing.company_id,
filing.title, filing.type, filing.date, filing.url,
filing.retrieval_date, filing.file_path, filing.processed_on,
filing.reference_date, filing.reference_quarter, company.id,
company.name, company.short_name, company.company_type,
company.market_type, company.last_update FROM (SELECT me.id,
me.filing_id, me.file_path, me.queued_on, me.processed_on,
me.processing FROM filing_pdf me JOIN filing filing ON filing.id =
me.filing_id JOIN company company ON company.id = filing.company_id
WHERE ( ( me.file_path IS NULL AND processing = ? AND ticker IS NOT
NULL ) ) GROUP BY me.id) me JOIN filing filing ON filing.id =
me.filing_id JOIN company company ON company.id = filing.company_id
WHERE ( ( me.file_path IS NULL AND processing = ? AND ticker IS NOT
NULL ) )
</sql>

The query generated by 0.08112 is not only much more convoluted, with
subselects, but outright wrong, as it does not join the "tickers"
relationship.

The relevant parts of the schema are below (column definitions
omitted, for brevity's sake):

<DBIC Schema>
package Filings::Schema::FilingPDF;
__PACKAGE__->load_components("InflateColumn::DateTime", "Core");
__PACKAGE__->table("filing_pdf");
__PACKAGE__->set_primary_key("id");
__PACKAGE__->belongs_to(
  "filing",
  "Filings::Schema::Filing",
  { id => "filing_id" },
);

package Filings::Schema::Company;
__PACKAGE__->load_components("InflateColumn::DateTime", "Core");
__PACKAGE__->table("company");
__PACKAGE__->set_primary_key("id");
__PACKAGE__->add_unique_constraint("name", ["name"]);
__PACKAGE__->has_many(
  "tickers",
  "Filings::Schema::CompanyTicker",
  { "foreign.company_id" => "self.id" },
);
__PACKAGE__->has_many(
  "filings",
  "Filings::Schema::Filing",
  { "foreign.company_id" => "self.id" },
);

package Filings::Schema::CompanyTicker;
__PACKAGE__->load_components("Core");
__PACKAGE__->table("company_ticker");
__PACKAGE__->set_primary_key("company_id", "ticker");
__PACKAGE__->add_unique_constraint("ticker", ["ticker"]);
__PACKAGE__->belongs_to(
  "company",
  "Filings::Schema::Company",
  { id => "company_id" },
);
</DBIC Schema>

Any suggestions of sane work-arounds?
Is this a bug or a feature?

For now I think I'm downgrading to a previous version which works.
Might as well try to figure out when it stopped working exactly
between 0.08107 and  0.08112 (if not on the latest version).

-Nilson Santos F. Jr.