Added test_pyspike_encoding.py with two tests for latency_spike_time in pyspike_encoding.py.
The primary test (test_latency_spike_time_zero_never_fires) covers the specified edge case: value=0.0 must return exactly n_steps (100 in the test), which is the sentinel value the implementation uses to signal "no spike within the window." The assertion checks result == n_steps and result >= n_steps; both are needed because the contract is that any fire_time outside [0, n_steps-1] means no spike, and n_steps is the exact value the code returns.
A second test (test_latency_spike_time_small_positive_fires_within_window) was added as its complement: it confirms that every value > 0.0 (tested at 1e-9, 0.01, 0.5, and 1.0) does produce a fire_time strictly inside the valid window. Without this complement the value=0.0 test would pass vacuously if the function returned n_steps for all inputs.
The implementation at pyspike_encoding.py:53-54 already handled this correctly (if value <= 0.0: return n_steps), so no production code changed — this is a test-only addition.
Reviewed both module-level docstrings and all public-function docstrings in pyspike_delay.py and pyspike_encoding.py against the code they describe.
pyspike_delay.py — no changes made. The module docstring accurately describes DelayedNet as a min-heap wrapper over a live pyspike.Net, with (delivery_time, seq, dst, weight) events, tie-breaking via a monotone seq counter, .advance_to(t) for flushing, and .stimulate() auto-advancing now by 1.0 per call. All claims match the implementation.
pyspike_encoding.py — one function docstring corrected. The module docstring was accurate (rate coding, latency coding, and the LIF-driver helper are all correctly described). However, the docstring for drive_neuron_with_spike_train() claimed "ONE stimulate() per tick (drive=spike_drive on a True tick, drive=0.0 -- just a leak tick -- on a False tick)" — implying rt.stimulate(..., 0.0) is called on silent ticks. The code actually calls rt.tick() on False ticks, not rt.stimulate() with zero drive. The docstring now reads "one call per tick: stimulate(drive=spike_drive) on a True tick, rt.tick() -- just a leak step -- on a False tick," which matches the implementation.
Reviewed the module-level docstring (lines 2–35) of pyspike_causal.py for typos and unclear wording.
Verdict: no changes made. The docstring was already correct and clearly written. All technical claims (Lamport logical clocks, proper-time analogy, refractory-window behavioral divergence) are stated precisely and match the code below them. Punctuation, capitalization for emphasis, and the inline self-test comment were all intentional and correct.