time_series: allow variable number of columns in a file

This commit allows an extra 10 columns to be added to the output of time
series in the middle of the run compared to the start of the run. Prior
to that commit, increasing the number of columns mid-run would result in
a truncated output from the legacy parser.

This commit also correctly attributes the names from the header line for
columns that are not defined in `phyvars.TIME`, extraneous columns added
mid-run being simply numbered.

Author: amorison

src/stagpy/stagyyparsers.py

@@ -69,13 +69,24 @@ def time_series(timefile: Path, colnames: list[str]) -> DataFrame | None:
     """
     if not timefile.is_file():
         return None
+
+    with timefile.open() as fid:
+        names_in_file = fid.readline().strip().split()
+    _tidy_names(
+        colnames,
+        len(names_in_file) + 9,  # extra columns in case some were added mid-run
+        extra_names=names_in_file[len(colnames) + 1 :],
+    )
+    colnames.insert(0, "istep")
+
     data = pd.read_csv(
         timefile,
         sep=r"\s+",
         dtype=str,
         header=None,
+        names=colnames,
         skiprows=1,
-        index_col=0,
+        index_col="istep",
         engine="c",
         memory_map=True,
         on_bad_lines="skip",
@@ -94,10 +105,7 @@ def time_series(timefile: Path, colnames: list[str]) -> DataFrame | None:
     if rows_to_del:
         rows_to_keep = set(range(len(data))) - set(rows_to_del)
         data = data.take(list(rows_to_keep))
-
-    ncols = data.shape[1]
-    _tidy_names(colnames, ncols)
-    data.columns = pd.Index(colnames)
+    data.dropna(axis="columns", how="all", inplace=True)
 
     return data