Deal with the recordings, not just read them

A night's scan leaves hundreds of files, most worth nothing and a few of
them the reason it was left running.  Sorting that out meant leaving the
browser and going at the directory with mv and rm.

Five keys, meant to be pressed once each going down the list:

  S I N   file it into saved/, investigate/ or noise/
  u       put the last one filed back
  d       delete it and its sidecars, for good -- asks first
  m       lock the frequency out, so no later scan stops on it

Each of these acts on the whole capture -- the .wav, the JSON sidecar, the
IQ, the transcript and the decoded data -- because a recording in one
directory and its transcript in another is a pair nothing will ever put
back together.  A move that cannot be finished puts back whatever already
moved.  The cursor stays on the row it was on, which is now the next
recording, since a cursor that jumped would make one-key-per-recording
impossible.

m writes to the lock-out list in the settings file, the same one the
scanner's own l key maintains, so a birdie found while reading last night's
recordings is gone from tonight's.  It says "the next scan": one already
running read its settings when it started.

The subdirectories sit under the recordings directory, so a scan writing
there never looks in them, and saunterbrowse ~/bandsaunter/saved reads one
back.

Also here, because this is the first part of the browser that writes:

 - The help screen is back inside eighty by twenty-four.  It had grown past
   the bottom of an ordinary window, which puts "q quit" off the screen.
 - The footer drops keys in a deliberate order when the window is narrow,
   rather than ellipsising whichever happened to be at the end.
 - Moving or deleting what is playing stops the player first.
 - The pty harness accepted an env and ignored it, so a test aimed at a
   throwaway settings directory wrote to the real one.  It honours it now,
   and conftest redirects the settings directory for every test besides.

1014 tests.

Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_016PsWPTweCT6pwxKngvVxcg
This commit is contained in:
The Dust Council 2026-08-29 13:21:32 -07:00
parent d6ae6d0c22
commit f9f0d94000
11 changed files with 1162 additions and 44 deletions

View file

@ -22,6 +22,12 @@ class Terminal:
"""One program running in a pty, with a screen that can be read back."""
def __init__(self, argv, cols=100, rows=30, env=None):
"""``env`` is added to the child's environment, not substituted for it.
It was accepted and silently ignored until a test that pointed the
program at a throwaway settings directory wrote to the real one
instead.
"""
import pyte
self.cols, self.rows = cols, rows
@ -35,7 +41,7 @@ class Terminal:
self.pid, self.fd = pty.fork()
if self.pid == 0: # pragma: no cover -- the child
child = dict(os.environ, TERM="xterm-256color",
PYTHONUNBUFFERED="1")
PYTHONUNBUFFERED="1", **(env or {}))
# rich reads COLUMNS and LINES in preference to asking the
# terminal, so leaving them set would make every resize invisible
# and every one of these tests pass for the wrong reason.