vendor: three %-formatted __repr__ methods keep the f-string disable alive in ftp.command and http.method

[JarryShaw]

Describe the bug

#803 finishes #798's sweep but cannot drop # pylint: disable=consider-using-f-string from two files, because three __repr__ methods still format with %. Verified at #803's head (0a5476491):

pcapkit/const/ftp/command.py:40    return "<%s [%s]>" % (self.__class__.__name__, self._name_)
pcapkit/const/ftp/command.py:119   return "<%s.%s: %s>" % (self.__class__.__name__, self._name_, self.desc)
pcapkit/const/http/method.py:42    return "<%s.%s>" % (self.__class__.__name__, self._value_)

Per-file state after #803 — the disable is earned in exactly the two files that still have %:

const/ftp/command.py          %-format lines=2  disable=1
const/http/method.py          %-format lines=1  disable=1
const/tcp/flags.py            %-format lines=0  disable=0   <- dropped by #803
const/reg/apptype/apptype.py  %-format lines=0  disable=0   <- dropped by #803
vendor/ftp/command.py         %-format lines=3  disable=2
vendor/http/method.py         %-format lines=2  disable=2

Expected behavior

f-strings in all three __repr__s, and the disable dropped from both files and both generators — at which point nothing in pcapkit/{const,vendor}/{ftp,http}/ needs it.

Additional context

These were out of #798's scope, not overlooked. #798 named "three %-formatted dunders in the AppType template" specifically, so #803 correctly stopped there and pinned the reason in a test docstring rather than silently leaving a stale disable. This issue is the remainder, filed so the disable's survival is tracked rather than inferred.

The maintainer's convention from #783: "id like to keep f-string convention across the library. only use % substitution when inevitable." A __repr__ is not an inevitable case — AppType.__repr__ was converted by #803 with member-by-member byte-identity proved over all 12,391 members, so the same proof shape applies here at far smaller scale.

Three traps, all measured on #798/#803:

• These const files are generated. Edit pcapkit/vendor/** and pcapkit/const/**. In the vendor templates braces in emitted code must be doubled, since BASE is a lambda returning an f-string, not a str.format() template.
• Do not regenerate anything under reg/apptype/ — AppType.LINK fetches IANA's live CSV. These two files do not crawl IANA, but prove equivalence the same way: render the template and diff the changed region, expecting 0 diff lines elsewhere.
• A %-format grep must allow both quote styles. Mine used '[^']*%[sdr] and reported 0 for these very files, because every surviving line is double-quoted. Self-test the pattern against a known-positive before trusting a zero.

Related: #798, #803, #783, #796.

[JarryShaw]

Checkable blocker: #803 merged.

gh pr view 803 -R JarryShaw/PyPCAPKit --json state,mergedAt

All four files this touches are files #803 is currently rewriting — pcapkit/{const,vendor}/{ftp/command,http/method}.py. Landing this first means the later of the two rewrites silently discards the other's edits, with no conflict marker to show it happened. #803 is the larger change and is already up, so it goes first.

Unblocks the moment #803 merges, at which point the whole remaining scope is three __repr__ lines plus two disable comments.

[JarryShaw]

#803 merged, so the original blocker has cleared — but this now waits on #796 instead. Re-blocked rather than opened.

Checkable blocker: #796 merged.

gh issue view 796 -R JarryShaw/PyPCAPKit --json state

#796 rewrites thirteen re.sub sites across pcapkit/vendor/, two of which are files this issue also edits — vendor/http/method.py and vendor/tcp/flags.py. A worker is on #796 now. Landing this first means one rewrite silently discards the other's edits.

Note #803 already dropped the disable from const/tcp/flags.py and const/reg/apptype/apptype.py, so the remaining scope here is exactly the three %-formatted __repr__s in const/ftp/command.py and const/http/method.py plus their two generators — re-derive the line numbers against origin/main (4530424df), since #803 moved them.

[JarryShaw]

Unblocked — #796 merged via #813 at 22:26Z. main is now 55e1b756e, so pcapkit/vendor/http/method.py and vendor/tcp/flags.py are free.

Dispatching a worker. No overlap with #815 (the only open PR touching const/), verified by path intersection — #815 is confined to const/reg/apptype/**, vendor/reg/apptype/apptype.py and foundation/registry/protocols.py.

Scope is unchanged and small: the three %-formatted __repr__s keeping the disable alive, plus the two generators. Re-derive every line number against origin/main (55e1b756e) — #803 and now #813 have both rewritten these files since the citations in this thread were written.

One correction to carry across from #813's review: my #796 text claimed the %-vs-flags damage was visible in the generated output. It is not — Vendor.wrap_comment() uses textwrap.wrap, whose replace_whitespace=True masks it. That does not change this issue's scope, but it is the same class of claim, so verify the __repr__ conversions change emitted output only where intended rather than assuming they are inert.