Correct histograms for zero byte allocations

If the program performs zero byte allocations, pretend that they were
one byte allocations instead. This way they'll be counted in the lowest
histogram bucket, as they should be, but we won't attempt to call
`math.log(0)` and get a math domain error.

Signed-off-by: Matt Wozniski <[email protected]>

src/pytest_memray/plugin.py

@@ -85,12 +85,10 @@ def histogram(
 
 def cli_hist(data: Iterable[float], bins: int, *, log_scale: bool = True) -> str:
     bars = " ▁▂▃▄▅▆▇█"
+    if log_scale:
+        data = [math.log(number if number else 1) for number in data]
     low = min(data)
     high = max(data)
-    if log_scale:
-        data = map(math.log, filter(lambda number: number != 0, data))
-        low = math.log(low)
-        high = math.log(high)
     data_bins = histogram(data, low=low, high=high, bins=bins)
     bar_indexes = (int(elem * (len(bars) - 1) / max(data_bins)) for elem in data_bins)
     result = " ".join(bars[bar_index] for bar_index in bar_indexes)

tests/test_pytest_memray.py

@@ -660,6 +660,7 @@ def test_memory_alloc_fails():
 @pytest.mark.parametrize(
     "size, outcome",
     [
+        (0, ExitCode.OK),
         (1, ExitCode.OK),
         (1024 * 1 / 10, ExitCode.OK),
         (1024 * 1, ExitCode.TESTS_FAILED),