rec_start

Start a continuous real-time recording of online PLC variables (runs in the background until rec_stop, duration_s, max_events or stop_when). Select by explicit symbols (an FB or struct instance expands to all its leaves), prefix ('MAIN.fbFill', 'GVL_Sim.aCh[1]' → every leaf under it) or pattern — the same leaves rec_scan lists for it: ''/'?' wildcards, '[ ]' literal ('GVL_Sim.aCh[1].'), a leaf is recorded when it or one of its parents matches ('*bOpen' → every bOpen). mode: 'cycle' = one PLC-timestamped sample per symbol per cycle_ms (ADS notification, cycle-accurate); 'change' = PLC-timestamped sample only when the value changes (checked every cycle_ms); 'poll' = sum-read every cycle_ms from this PC. Triggers: start_when='MAIN.bStart == TRUE' with pre_trigger_ms keeps the events before the trigger (scope-style); stop_when='MAIN.fbFill.bDone == TRUE'. Events stream to a JSONL file. duration_s is PLC time from the first sample (also while waiting for start_when); post_trigger_s ends the recording that long after the start_when trigger fired (scope-style post-trigger window). Timestamps (cycle/change) are the TwinCAT runtime's own, stored exactly with the raw bytes it sent; neighbouring variables share ONE notification (a block), which is what keeps 1 ms recordings lossless — select whole FBs, structs or arrays rather than hundreds of scattered globals. Any notification that cannot be registered fails the start (allow_partial=true records the rest and lists unsubscribed_symbols). max_delay_ms > 0 lets the runtime batch samples (fewer ADS frames; each sample keeps its own timestamp). cycle mode also records a per-cycle counter as ground truth (cycle_counter='auto' = the matching task's _TaskInfo[n].CycleCount, or your own e.g. 'MAIN.nCycleCount', or null for none) and the header carries a manifest of what was recorded from (AppTimestamp, OnlineChangeCnt, tasks, TwinCAT build, symbol version). A recording is of one program: a download / online change while it runs ends it at the last sample verified against the symbol version (program_changed); nothing sampled from the new program is written. Conditions are levels: a start_when that already holds fires at once — edge_triggers=true fires only when it BECOMES true. A condition may name one element of an array recorded as a whole (GVL_Test.aCounter[1] when GVL_Test.aCounter is recorded). engine (cycle mode): 'native' = the C++ capture service (Beckhoff AdsLib on TcAdsDll: ADS callback → ring → file, no Python on the data path), 'python' = the in-process recorder, None = recorder.engine from the config (default 'auto': native when installed). The response says which engine ran and why. After stopping: rec_verify. also_scope: also run the TwinCAT Scope configuration open in XAE (scope = its name when several are open) — StartRecord now; at rec_stop StopRecord and SaveSVD into /Scope Recordings/.svdx, the CSV .scope.csv exported from that file with Beckhoff's export tool (XAE's ExportCSV only without it, flagged), XAE's ExportCSV of what Scope holds in memory beside it (.scope.xae.csv) with both row counts compared (cross_check — a saved .svdx shorter than memory, or the reverse, is a warning), a signed link file, and the Scope View log's new ERROR / FATAL lines (scope_view_log). A .svdx still being written 15 s after SaveSVD is not exported. What Scope records (channels, sample time) is set in that configuration. With recorder.also_scope = 'ask' (default) and a Scope configuration open, leaving also_scope unset asks the user (prompt, or an error that tells you to ask and call again with true/false). After a recording Scope holds its data until it is disconnected: a recording with Scope then asks to disconnect it first — scope_disconnect=true / false is the user's answer (false: this recording without Scope).