A column declared with `NOT NULL DEFAULT ''` apply-looped: every
post-apply plan re-emitted `MODIFY COLUMN … DEFAULT ''` because the
catalog returned the bare empty string from
`information_schema.COLUMNS.COLUMN_DEFAULT` while the parser side
stored `''` (quoted empty string). They never compared equal.

`catalog.normalizeColumnDefault` short-circuited on `def == ""`
before its vitess-based bareword check, which was the root cause —
removing the short-circuit alone doesn't help because vitess can't
parse `SELECT` with an empty expression. The fix is type-aware:
when the column type accepts `DEFAULT ''` as the bare empty string
(VARCHAR / CHAR / ENUM / SET), normalise to `''`. The function
finally consumes its long-dormant `typeName` argument.

BINARY / VARBINARY have a different round-trip drift (MySQL stores
their `''` default as a hex literal `0x` / `0x000000…`). That needs
its own catalog-side normalisation; filed in TODO.md as a follow-up.

Tests
  - catalog/catalog_test.go: TestColumnDefaultEmptyStringNormalisation
    sets up a table with VARCHAR / CHAR / ENUM / SET DEFAULT ''
    columns and asserts each normalises to `''`.
  - testdata/apply/empty_string_default_no_drift.yml: end-to-end
    YAML fixture with `verify_no_drift: true` so the round-trip is
    pinned through the full plan/apply pipeline.
  - test/scenario/testdata/evolution/0[2-5]_*.sql: restore the
    `NOT NULL DEFAULT ''` clauses on display_name (PR #31 worked
    around the bug by dropping the explicit DEFAULT). The evolution
    scenario now exercises the empty-default round-trip too.
  - test/scenario/evolution.test.sh: header comment updated.

TODO.md
  - Remove the empty-DEFAULT entry (fixed here).
  - Add a follow-up entry for the BINARY / VARBINARY case so the
    workaround is documented.

Verified on both compose services (mysql 8.0, mysql 9.4): full Go
suite, YAML harness, scenario suite all pass.

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>

Three inline comments, plus a refinement uncovered while writing the
new test:

catalog/tables.go + catalog/catalog_test.go (typography fix)
  - Replace stray U+201D ("right double quotation mark") with the
    ASCII `''` literal in the doc comments. Editor / clipboard
    contamination from the previous round; the SQL examples now
    read correctly. (Copilot #1, #2.)

catalog/tables.go (BINARY/VARBINARY split)
  - The original `columnTypeAllowsEmptyStringDefault` excluded
    both BINARY and VARBINARY together. Verifying with information_
    schema showed only fixed-width BINARY surfaces its empty
    default as a hex literal (`0x` for the degenerate case;
    `0x000000…` for non-zero N). VARBINARY (variable-length)
    surfaces the bare empty string just like VARCHAR, so it
    round-trips cleanly via the empty-string path. Narrow the
    exclusion to BINARY only and add VARBINARY to the supported
    list. The function now checks `varbinary` before `binary` so
    HasPrefix doesn't let one match the other.

catalog/catalog_test.go (test for the BINARY exclusion)
  - TestColumnDefaultBinaryEmptyStringIsHex pins the documented
    BINARY behaviour and catches a future refactor that
    accidentally normalises the hex literal to ''. The companion
    TestColumnDefaultEmptyStringNormalisation gains a VARBINARY
    column to exercise the type's now-supported path. (Copilot #3.)

`make test` (8.0) — pass; `make test-scenario` (8.0) — pass.

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>

Four Copilot inline comments on PR #33:

catalog/tables.go (doc) + TODO.md
  - Doc and TODO still listed VARBINARY alongside BINARY as "needs
    its own normalisation". The previous round actually moved
    VARBINARY into the supported set (it surfaces its DEFAULT '' as
    the bare empty string, so the existing path round-trips it
    cleanly). Update the function-doc and TODO to spell out that
    only fixed-width BINARY remains broken. (Copilot #1, #2.)

testdata/apply/empty_string_default_no_drift.yml
  - Add a VARBINARY(8) column to the no-drift fixture so the new
    branch is exercised through plan + apply + drift-check, not
    just the catalog unit test. (Copilot #3.)

catalog/tables.go + catalog/catalog_test.go (typography)
  - Bulk replace stray U+201D ("right double quotation mark") with
    ASCII `''` across both files. The previous round's Edit tool
    ate the quote characters; verified the bytes with a follow-up
    grep. (Copilot #4.)

`make test` (8.0) — pass; `make test-scenario` (8.0) — pass.

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>

`misspell` was rewriting consecutive ASCII single-quotes (`''`) into a
typographic right-double-quote in comments, which then became a Copilot
review nit. Drop the linter — the project is small enough that an
in-band misspelling check isn't paying for itself.

Independently, `gofmt` in Go 1.26 also collapses `''` to `”` in
comments, so even with misspell gone the literal pair can't survive a
format pass. Rewrite the empty-default explanation in tables.go and
catalog_test.go to refer to "the quoted empty literal" / "a pair of
single-quote characters" in prose instead, which both readers and
gofmt are fine with.

Also fold in the function-body restructure for
`columnTypeAllowsEmptyStringDefault`: BINARY is the sole excluded
type, so check it first and let the supported list (VARCHAR / CHAR /
VARBINARY / ENUM / SET) be a flat OR — the previous "VARBINARY before
BINARY" ordering hinted at a HasPrefix collision that doesn't actually
exist (HasPrefix("varbinary", "binary") is false).

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>