mark-twin 0.5.0 → 0.6.0

checksums.yaml

@@ -1,7 +1,7 @@
 ---
 SHA256:
-  metadata.gz: d54022c14550637480eed5ecea2c7cf5b391d75724559c5c25cdac8cef134526
-  data.tar.gz: c5fc90db5a080109b60f330f591cb62055412ca3cf59f7d68d89e156b6d17f85
+  metadata.gz: 9e643b14b2fe0dd153539ff16d10a0950f3c478d327620dec1b9b04d7af4f340
+  data.tar.gz: 4eb24c0280ee5f1085024c1357a6d7b8dd0b7f16198e3cce8b4c9ce5849705b5
 SHA512:
-  metadata.gz: 4511a3d7293d69be4cabd081c772ac252a0123f4207cabd007da52186b3a4cc0f7edff487b8530484ab083ecab0ced150060c3912c4605021689ad3fc9e5caf1
-  data.tar.gz: c4968f1c11bebaa8b7b883f9271e5b55495a8bde93f6557d8a5e2842bac5c263c8e859c0a74e35ee4b119f3d75056170a0016d4f3516804c1926d136709a62f8
+  metadata.gz: 7d5567dc0637bc767f7ddff5716a2e5031902e787958e1ecd69f54931e737b2946d82c999637d6610d732e5063898c3b717b49148f8e495af328ea2f0b716d03
+  data.tar.gz: 94df8bd72ea5548e668b1173e17b9e9bf54efc1fbca4576c109ed64740794decb94eb23fab78c50ed2b0654bc2160bb1a4be85c74a2f9281bd2a178e0f33b270

data/ARCHITECTURE.md

@@ -53,7 +53,7 @@ test/test_pure.rb
 
 ```
 program, path, description, active, excludes, owned, label, source, target, cmd,
-delete, render, render_outdated, target_path_field, sync_file,
+delete, render, render_outdated, target_path_field, sync_file, verify,
 source_exists, target_exists, source_mtime, target_mtime, conflict,
 directory, content_equal, drift
 ```
@@ -68,7 +68,10 @@ content-verified when set. Directory jobs get no mtime judgment at all: a
 directory's mtime moves on every sync and on every excluded file, so
 `Job#status` reports `unverified` until `twin status` fills `drift` by asking
 rsync (`Conflict.drift` — paired dry-runs, itemize classification, checksums
-for timestamp-only candidates).
+for timestamp-only candidates). `Verify: false` opts a job out of every
+content round — scanner md5, drift dry-runs, pre-sync conflict detection —
+for entries where the walk itself is the cost; its status stays mtime-based
+(files) or `unverified` (directories).
 
 `Job#status` → one of `disabled / unreachable / both_missing / missing_source /
 missing_target / target_newer / in_sync / source_newer / unverified`. Render
@@ -164,16 +167,28 @@ rationale.
 Two stages:
 
 1. **Stage 1 — program picker.** Multi-line NUL-separated entries (`--read0`).
-   Each entry has a header (icon, program name, job count, sync-file) and
+   Each entry has a header (icon, program name, job count, sync-files) and
    indented body lines (one per job). No preview. Single-select. ESC exits.
+   Same-named programs (case-insensitive) from different sync-files are merged
+   into one entry, and a trailing bracket group joins by convention:
+   `livesync [agent]` merges under `livesync`. A picker-only view
+   (`Picker.merge_programs`); the data model, `status`, `sync -p` and JSON
+   keep the per-file programs. Jobs stay grouped per file in document order,
+   so a file's closing `Cmd` block still fires after its own paths.
 2. **Stage 2 — path multi-picker.** Tab-delimited rows (`id\tdisplay`),
-   `--with-nth=2` hides the `id`. Multi-select via Tab. Preview pane shows
-   the *compact* view (frontmatter + intro + the heading section containing
-   the YAML block for the highlighted path), rendered via
+   `--with-nth=2` hides the `id`. Multi-select via Tab. On open, directory
+   jobs get their drift verified (`Conflict.fill_drift` — affordable for one
+   program, unlike for the whole of stage 1). A merged program's rows are
+   sectioned per sync-file by dim header rows whose id is `-`; they map to no
+   job, so toggling one is inert. Preview pane shows the *compact* view
+   (frontmatter + intro + the heading section containing the YAML block for
+   the highlighted path) from the job's own sync-file, rendered via
    `apex --plugins -t terminal256`. ESC returns to Stage 1.
 
 Compact previews are pre-rendered to per-job tempfiles before fzf launches.
-An `awk` lookup maps `{1}` (the id) → tempfile path.
+An `awk` lookup maps `{1}` (the id) → tempfile path. The Tempfile objects
+stay referenced while fzf runs — a GC'd Tempfile unlinks its file under the
+running preview.
 
 ## CLI
 

data/README.md

@@ -58,7 +58,10 @@ on both sides since the last sync, and paths that differ per host.
 ## Screenshots
 
 Stage 1 — program picker. One row per program, color-coded status, indented
-paths underneath:
+paths underneath. A name that appears in several sync-files
+(the app in one, its config in another) is one entry, and a trailing bracket
+group ties variants to their base name — `livesync [agent]` lists under
+`livesync`. Stage 2 sections the merged paths per origin:
 
 ![Stage 1 — program picker](https://raw.githubusercontent.com/rhsev/mark-twin/main/docs/stage_1.png)
 
@@ -213,7 +216,7 @@ rather than as an error.
 
 | Field | Where | Meaning |
 |---|---|---|
-| `Program` | block | Group name; blocks sharing it sync together |
+| `Program` | block | Group name; blocks sharing it sync together. A trailing `[…]` lists under the base name in the picker |
 | `Path` | block | Path relative to `Source` (file or directory) |
 | `Source` | either | Absolute base path on this machine |
 | `Target` | either | Absolute base path, or `user@host:/path` for ssh |
@@ -226,6 +229,7 @@ rather than as an error.
 | `Delete` | block | `true` mirrors deletions, with backups |
 | `Cmd` | block | Shell command, run only when bytes actually moved |
 | `Render` | block | `true` substitutes `{{tokens}}` instead of copying |
+| `Verify` | block | `false` skips content verification — for entries too big (see below) |
 
 ### Exclude or Own?
 
@@ -376,9 +380,19 @@ Two properties make this bearable day to day:
 
 `twin status` gives the same verdict without syncing: it runs the dry-runs per
 entry and reports what would flow, what merely differs in timestamp, and what
-changed on the target. Only the picker stays mtime-blind for directory entries
-— it marks them `∘` (unverified) rather than guessing, because the dry-runs
-would make it slow to open.
+changed on the target. The picker's first stage stays mtime-blind for
+directory entries — it marks them `∘` (unverified) rather than guessing,
+because the dry-runs would make it slow to open. Entering a program runs them
+for just that program, so its directory rows show a real verdict.
+
+One escape hatch: `Verify: false` opts an entry out of every content round —
+the md5 checks, the status dry-runs, the picker's verification, the pre-sync
+conflict listing. It exists for entries where the walk itself is the cost: a
+`node_modules` tree on an SMB mount stats tens of thousands of files for one
+verdict. Such an entry stays `∘`/mtime-based, and the picker's second stage
+says so in its header rather than letting the `∘` pass as pending; a sync
+still leaves newer target files alone (`--update` holds), they just aren't
+itemised first.
 
 ## Automation
 
@@ -443,7 +457,7 @@ it differs from the current target, so a `Cmd` hook fires only on a real change.
 ## LiveSync LaunchAgent
 
 ```yaml
-Program: livesync-agent
+Program: livesync [agent]
 Source: "{{src.home}}/Automation/launchd"
 Path: com.ralf.livesync.plist
 Target: "{{dst.mount}}"

data/lib/twin/cli.rb

@@ -93,7 +93,7 @@ module Twin
         exit 1
       end
 
-      programs = Scanner.load_programs(cfg, file: file, show_all: false)
+      programs = Picker.merge_programs(Scanner.load_programs(cfg, file: file, show_all: false))
       if programs.empty?
         warn "no active programs found#{" in #{file}" if file}"
         return
@@ -125,7 +125,7 @@ module Twin
         break if $stdin.gets&.strip == "q"
 
         # reload so status reflects what was just synced, stay on this program
-        programs     = Scanner.load_programs(cfg, file: file, show_all: false)
+        programs     = Picker.merge_programs(Scanner.load_programs(cfg, file: file, show_all: false))
         selected_key = [program.name, program.sync_file]
       end
     end
@@ -168,7 +168,11 @@ module Twin
           puts "    #{j.path}#{conflict}"
           if j.directory
             # Directory mtimes prove nothing — the drift verdict replaces them.
-            puts "      #{j.drift ? drift_summary(j.drift) : "(not checked — target unavailable)"}"
+            note = if j.drift then drift_summary(j.drift)
+                   elsif j.verify == false then "(not checked — Verify: false)"
+                   else "(not checked — target unavailable)"
+                   end
+            puts "      #{note}"
           else
             src = j.source_exists ? j.source_mtime.strftime("%Y-%m-%d %H:%M:%S") : "(not found)"
             tgt = j.target_exists ? j.target_mtime.strftime("%Y-%m-%d %H:%M:%S") : "(not found)"
@@ -183,30 +187,9 @@ module Twin
       end
     end
 
-    # Ask rsync what a sync of each directory job would actually do — the
-    # dry-runs are subprocess I/O, so a few run in parallel. File jobs are
-    # already content-checked by the scanner and need no second look.
+    # Ask rsync what a sync of each directory job would actually do.
     def verify_drift(cfg, programs)
-      jobs = programs.flat_map(&:jobs).select do |j|
-        j.directory && j.active == 1 && !j.target_unreachable &&
-          j.source_exists && j.target_exists
-      end
-      return if jobs.empty?
-
-      queue = Queue.new
-      jobs.each { |j| queue << j }
-      Array.new([4, jobs.size].min) do
-        Thread.new do
-          loop do
-            j = begin
-              queue.pop(true)
-            rescue ThreadError
-              break
-            end
-            j.drift = Twin::Conflict.drift(cfg, j)
-          end
-        end
-      end.each(&:join)
+      Twin::Conflict.fill_drift(cfg, programs.flat_map(&:jobs))
     end
 
     def drift_summary(d)
@@ -314,7 +297,10 @@ module Twin
       return true  if force
       return false if skip_conflicts || dry_run
 
-      conflicts = jobs.flat_map { |j| Twin::Conflict.detect(cfg, j) }
+      # Verify: false jobs skip the detection round (too big to walk); rsync's
+      # --update still keeps newer target files, they just aren't listed here.
+      conflicts = jobs.reject { |j| j.verify == false }
+                      .flat_map { |j| Twin::Conflict.detect(cfg, j) }
       return false if conflicts.empty?
 
       report_conflicts(conflicts)

data/lib/twin/conflict.rb

@@ -79,6 +79,38 @@ module Twin
       drift(cfg, job)&.conflicts || []
     end
 
+    # Directory jobs whose drift is still unknown and answerable — both sides
+    # present, target reachable, not opted out via Verify: false. File jobs
+    # are content-checked by the scanner and need no second look.
+    def drift_candidates(jobs)
+      jobs.select do |j|
+        j.drift.nil? && j.verify != false && j.directory && j.active == 1 &&
+          !j.target_unreachable && j.source_exists && j.target_exists
+      end
+    end
+
+    # Fill `drift` on every candidate by asking rsync — the dry-runs are
+    # subprocess I/O, so a few run in parallel.
+    def fill_drift(cfg, jobs)
+      jobs = drift_candidates(jobs)
+      return if jobs.empty?
+
+      queue = Queue.new
+      jobs.each { |j| queue << j }
+      Array.new([4, jobs.size].min) do
+        Thread.new do
+          loop do
+            j = begin
+              queue.pop(true)
+            rescue ThreadError
+              break
+            end
+            j.drift = drift(cfg, j)
+          end
+        end
+      end.each(&:join)
+    end
+
     # Pure assembly of a Drift from parsed entries. A file the normal run
     # would also transfer flows source→target as intended; one only the
     # forced run would touch is being held back by --update — the target owns

data/lib/twin/picker.rb

@@ -2,6 +2,7 @@ require "open3"
 require "tempfile"
 
 require_relative "preview"
+require_relative "conflict"
 
 module Twin
   # Two-step fzf picker:
@@ -14,9 +15,10 @@ module Twin
       source_newer:   "→",
       target_newer:   "←",
       in_sync:        "✓",
-      # Directory jobs in the picker: their mtimes prove nothing, and the
-      # dry-run that would prove something is too slow for a picker to open
-      # with. No claim instead of a wrong one — `twin status` has the verdict.
+      # Directory jobs in stage 1: their mtimes prove nothing, and the
+      # dry-runs that would prove something are too slow for the full picker.
+      # No claim instead of a wrong one — stage 2 verifies its one program
+      # on open, and `twin status` has the full verdict.
       unverified:     "∘",
       missing_target: "!",
       missing_source: "!",
@@ -49,6 +51,29 @@ module Twin
 
     # ── Stufe 1: program picker ───────────────────────────────────────────────
 
+    # A trailing bracket group ties a program to its base name by convention:
+    # "livesync [agent]" lists under "livesync". Only the suffix form counts —
+    # brackets elsewhere are just a name.
+    BRACKET_SUFFIX = /\s*\[[^\]]*\]\z/
+
+    def base_name(name) = name.sub(BRACKET_SUFFIX, "").strip
+
+    # One stage-1 entry per base name (case-insensitive), even when it appears
+    # in several sync-files or as bracketed variants — "everything of
+    # fileview" is one decision. The list is sorted by name; document order
+    # only orders the *jobs*, program order carries no meaning. Purely a
+    # picker view: the data model, `status`, `sync -p` and JSON keep the
+    # per-file programs. Jobs stay grouped per file in document order, so a
+    # file's closing Cmd block still fires after its own paths.
+    def merge_programs(programs)
+      programs.group_by { |p| base_name(p.name).downcase }.values.map do |group|
+        next group.first if group.size == 1
+        name = group.find { |p| !p.name.match?(BRACKET_SUFFIX) }&.name ||
+               base_name(group.first.name)
+        Program.new(name: name, jobs: group.flat_map(&:jobs))
+      end.sort_by { |p| p.name.downcase }
+    end
+
     # Multi-line entries (NUL-separated). Header line per program, indented
     # body lines per job. Returns the selected Program or nil.
     def pick_program(programs)
@@ -90,14 +115,45 @@ module Twin
       jobs = program.jobs
       return [] if jobs.empty?
 
+      # Stage 2 covers one program, so the dry-runs that are too slow for the
+      # whole picker are affordable here — directory rows get a real verdict
+      # instead of `∘`. Already-verified jobs are skipped on re-entry.
+      if Conflict.drift_candidates(jobs).any?
+        $stderr.print dim("verifying directories …")
+        Conflict.fill_drift(cfg, jobs)
+        $stderr.print "\r\e[K"
+      end
+
       path_width = jobs.map { |j| j.path.length }.max
       tempfiles  = write_compact_previews(program, jobs)
       preview_cmd, mapfile = build_apex_preview_cmd(tempfiles, cfg)
 
-      rows = jobs.each_with_index.map do |j, i|
-        icon  = STATUS_ICONS[j.status] || "?"
-        line  = "#{icon}  #{j.path.ljust(path_width)}  #{job_delta(j)}"
-        "#{i}\t#{colorize(j.status, line)}"
+      # A merged program spans sync-files (and bracketed variants); a dim
+      # section row marks each block, named by its file — plus the original
+      # program name when it differs from the merged one. Section rows carry
+      # "-" instead of an index — fzf can't make them unselectable, but they
+      # map to no job, so toggling one is inert.
+      section = ->(j) { [j.program, j.sync_file] }
+      multi = jobs.map(&section).uniq.size > 1
+      rows = []
+      prev = nil
+      jobs.each_with_index do |j, i|
+        if multi && section.(j) != prev
+          file  = File.basename(j.sync_file.to_s)
+          label = j.program == program.name ? file : "#{j.program} · #{file}"
+          rows << "-\t#{dim("[#{label}]")}"
+          prev = section.(j)
+        end
+        icon = STATUS_ICONS[j.status] || "?"
+        line = "#{icon}  #{j.path.ljust(path_width)}  #{job_delta(j)}"
+        rows << "#{i}\t#{colorize(j.status, line)}"
+      end
+
+      header = "#{program.name} — Tab toggles, Enter confirms"
+      unverified = jobs.select { |j| j.verify == false }
+      unless unverified.empty?
+        header += "\n\e[33m⚠ content checks off (Verify: false): " \
+                  "#{unverified.map(&:path).join(", ")}\e[0m"
       end
 
       fzf = [
@@ -105,7 +161,7 @@ module Twin
         "--multi", "--ansi",
         "--delimiter=\t", "--with-nth=2",
         "--prompt=#{program.name} > ",
-        "--header=#{program.name} — Tab toggles, Enter confirms",
+        "--header=#{header}",
         "--preview=#{preview_cmd}",
         "--preview-window=right:60%:wrap",
         "--height=100%", "--reverse",
@@ -113,45 +169,55 @@ module Twin
         "--color=bg+:-1,hl+:reverse",
       ]
 
-      output, status = Open3.capture2(SH_ENV, *fzf, stdin_data: rows.join("\n"))
-      return :back if status.exitstatus == 130   # ESC / Ctrl-C
-      return [] unless status.success?
-      return [] if output.strip.empty?
+      loop do
+        output, status = Open3.capture2(SH_ENV, *fzf, stdin_data: rows.join("\n"))
+        return :back if status.exitstatus == 130   # ESC / Ctrl-C
+        return [] unless status.success?
+        return [] if output.strip.empty?
 
-      output.lines.filter_map do |line|
-        idx = line.split("\t", 2).first&.to_i
-        jobs[idx] if idx
+        selected = output.lines.filter_map do |line|
+          idx = line.split("\t", 2).first
+          jobs[idx.to_i] if idx&.match?(/\A\d+\z/)
+        end
+        return selected unless selected.empty?
+        # Enter landed on a section header (or only headers were toggled):
+        # nothing real was accepted — reopen rather than read it as "exit".
       end
     ensure
-      tempfiles&.each_value { |path| File.unlink(path) rescue nil }
-      File.unlink(mapfile) rescue nil if mapfile
+      tempfiles&.each_value { |f| f.close! rescue nil }
+      mapfile&.close! rescue nil
     end
 
-    # Write per-job compact-preview markdown to tempfiles. Returns {idx => path}.
+    # Write per-job compact-preview markdown to tempfiles. Returns
+    # {idx => Tempfile}. The Tempfile objects (not just their paths) must stay
+    # referenced while fzf runs: a GC'd Tempfile unlinks its file, and the
+    # preview command would find nothing to render. Each job's excerpt comes
+    # from its own sync-file — a merged program spans several.
     def write_compact_previews(program, jobs)
       result = {}
       jobs.each_with_index do |job, i|
-        compact = Preview.extract_compact(program.sync_file, job.path)
+        compact = Preview.extract_compact(job.sync_file, job.path)
         f = Tempfile.new(["twin-#{i}-", ".md"])
         f.write(compact)
         f.close
-        result[i] = f.path
+        result[i] = f
       end
       result
     end
 
-    # Returns [preview_cmd, mapfile_path]; the caller unlinks the mapfile.
+    # Returns [preview_cmd, mapfile]; the caller holds the Tempfile until fzf
+    # is done (GC would unlink it) and closes it afterwards.
     def build_apex_preview_cmd(tempfiles, cfg)
       # Map idx → file via a small TSV, awk picks the right one for {1}.
       mapfile = Tempfile.new(["twin-map-", ".tsv"])
-      tempfiles.each { |i, path| mapfile.puts("#{i}\t#{path}") }
+      tempfiles.each { |i, f| mapfile.puts("#{i}\t#{f.path}") }
       mapfile.close
 
       render_cmd = pick_renderer(cfg)
 
       cmd = %(F=$(awk -v id={1} -F'\\t' '$1==id {print $2}' #{mapfile.path}); ) +
             %([ -n "$F" ] && #{render_cmd})
-      [cmd, mapfile.path]
+      [cmd, mapfile]
     end
 
     # Pick the first available markdown renderer.
@@ -182,11 +248,11 @@ module Twin
 
     def render_program_entry(program, name_width, path_width)
       icon      = colorize(program.status, STATUS_ICONS[program.status] || "?")
-      sync_file = File.basename(program.sync_file.to_s)
+      files     = program.jobs.map { |j| File.basename(j.sync_file.to_s) }.uniq.join(" · ")
       count     = program.active_jobs.size
       total     = program.jobs.size
       header    = "#{icon}  #{bold(program.name.ljust(name_width))}    " \
-                  "#{dim("(#{count}/#{total})")}    #{dim("[#{sync_file}]")}"
+                  "#{dim("(#{count}/#{total})")}    #{dim("[#{files}]")}"
 
       body = program.jobs.map do |j|
         icon  = STATUS_ICONS[j.status] || "?"

data/lib/twin/scanner.rb

@@ -10,7 +10,7 @@ module Twin
     :program, :path, :description, :active, :excludes, :owned, :label,
     :source, :target, :cmd, :delete, :render, :render_outdated, :target_path_field, :sync_file,
     :source_exists, :target_exists, :source_mtime, :target_mtime, :conflict,
-    :target_unreachable, :directory, :content_equal, :drift,
+    :target_unreachable, :directory, :content_equal, :drift, :verify,
     keyword_init: true,
   ) do
     def source_path = File.join(source, path)
@@ -129,7 +129,8 @@ module Twin
     # content clears the conflict — the timestamps merely disagree.
     def verify_remote_file_content(host, host_jobs)
       candidates = host_jobs.select do |j|
-        !j.directory && !j.render && j.source_exists && j.target_exists &&
+        j.verify != false && !j.directory && !j.render &&
+          j.source_exists && j.target_exists &&
           j.source_mtime && j.target_mtime && (j.target_mtime - j.source_mtime).abs >= 60
       end
       return if candidates.empty?
@@ -183,6 +184,11 @@ module Twin
       return nil if path.empty? || source.empty? || target.empty?
 
       render   = r["Render"] == true
+      # Verify: false — no content verification for this entry, ever: too big
+      # or too remote for md5/dry-run rounds (a node_modules tree over SMB).
+      # Status falls back to mtime for files and stays ∘ for directories; the
+      # pre-sync conflict check skips it (--update still protects the target).
+      verify   = r["Verify"] != false
       excludes = split_list(r["Exclude"])
       # Own: paths inside the sync scope that the TARGET owns — machine-specific
       # config the source must never clobber. Same rsync effect as Exclude, but
@@ -214,7 +220,7 @@ module Twin
       # (a `cat >` copy before the first twin run). Check before judging;
       # a directory's own mtime is judged not at all (see Job#status).
       content_equal = nil
-      if !render && !remote && !directory && src_exists && tgt_exists &&
+      if verify && !render && !remote && !directory && src_exists && tgt_exists &&
          src_mtime && tgt_mtime && (tgt_mtime - src_mtime).abs >= 60
         content_equal = Twin::Conflict.same_content?(src_full, tgt_full)
       end
@@ -248,6 +254,7 @@ module Twin
         target_unreachable: false,
         directory:        directory,
         content_equal:    content_equal,
+        verify:           verify,
       )
     end
 

data/lib/twin/version.rb

@@ -1,3 +1,3 @@
 module Twin
-  VERSION = "0.5.0"
+  VERSION = "0.6.0"
 end

metadata

@@ -1,7 +1,7 @@
 --- !ruby/object:Gem::Specification
 name: mark-twin
 version: !ruby/object:Gem::Version
-  version: 0.5.0
+  version: 0.6.0
 platform: ruby
 authors:
 - Ralf Hülsmann