Fix get_real_name for names with more than two dotted parts (#332)

[chuenchen309]

Fixes #332.

The bug

get_real_name() returns the wrong component for a fully-qualified name with more than two dotted parts:

import sqlparse
ident = sqlparse.parse("db.schema.tbl.col")[0].tokens[0]
ident.get_real_name()   # 'schema'  — should be 'col'
ident.get_name()        # 'schema'  — should be 'col' (get_name falls back to real name)

More cases on current master: a.b.c → 'b' (should be 'c'), x.y.z AS w → real name 'y' (should be 'z'), mydb.sch.func(1) → 'sch' (should be 'func'). Two-part names like a.b → 'b' are correct and stay correct.

This is issue #332 (open since 2017): the "object name" of db.schema.tbl.col is the last component, but the current result is an intermediate one that is neither the object nor its parent.

Root cause

NameAliasMixin.get_real_name (sqlparse/sql.py) anchors on the first dot:

dot_idx, _ = self.token_next_by(m=(T.Punctuation, '.'))   # first dot
return self._get_first_name(dot_idx, real_name=True)      # first name from there

token_next_by returns the first dot, so _get_first_name returns the component right after it — the second part. For a 2-part name the first dot is also the last, so it happened to be correct; for 3+ parts it stops too early.

get_parent_name documents the opposite intent — "A parent object is identified by the first occurring dot" — so the first-dot logic belongs there, not in get_real_name. The docstring of get_real_name says "the real name (object name)", which is unambiguously the component after the last dot.

The fix

Anchor get_real_name on the last dot instead of the first. get_parent_name is untouched, so parent name and real name now come from opposite ends of the dotted path, as intended.

Verification

• The repro above now returns col; a.b.c → c, x.y.z AS w → real z / alias w, mydb.sch.func(1) → func, db.sch."My Table" → My Table, a.b.* → *. Two-part names and get_parent_name are unchanged.
• Added test_get_real_name_multi_part_dotted; it fails on master (assert 'col' == 'schema') and passes with the fix.
• Full suite: 488 passed, 2 xfailed, 1 xpassed (was 487 + the new test) — no existing test relied on the old behavior. ruff check sqlparse/sql.py clean.

---

This PR was authored by an AI coding agent (Claude Code) running on this account: the AI found the bug, ran the repro, wrote the test, and wrote this description. The human account holder reviews every change and is accountable for it. The verification above is real and re-runnable from the diff. If this isn't the kind of contribution you want, say so and I'll close it.