From 7964dc85db49e85c7209f5aecf06f32e424361a9 Mon Sep 17 00:00:00 2001
From: Matthew Evans <git@ml-evs.science>
Date: Fri, 13 Jun 2025 15:49:23 +0100
Subject: [PATCH] Add mode to attempt to read files with unknown columns and
 only warn

---
 galvani/BioLogic.py | 77 +++++++++++++++++++++++++++++++++++++++------
 1 file changed, 68 insertions(+), 9 deletions(-)

diff --git a/galvani/BioLogic.py b/galvani/BioLogic.py
index 7c9d3d6..aa72045 100644
--- a/galvani/BioLogic.py
+++ b/galvani/BioLogic.py
@@ -13,9 +13,12 @@ from os import SEEK_SET
 import time
 from datetime import date, datetime, timedelta
 from collections import defaultdict, OrderedDict
+import warnings
 
 import numpy as np
 
+UNKNOWN_COLUMN_TYPE_HIERARCHY = ("<f8", "<f4", "<u4", "<u2", "<u1")
+
 
 def fieldname_to_dtype(fieldname):
     """Converts a column header from the MPT file into a tuple of
@@ -428,18 +431,23 @@ def parse_BioLogic_date(date_text):
     return date(tm.tm_year, tm.tm_mon, tm.tm_mday)
 
 
-def VMPdata_dtype_from_colIDs(colIDs):
+def VMPdata_dtype_from_colIDs(colIDs, error_on_unknown_column: bool = True):
     """Get a numpy record type from a list of column ID numbers.
 
     The binary layout of the data in the MPR file is described by the sequence
     of column ID numbers in the file header. This function converts that
-    sequence into a numpy dtype which can then be used to load data from the
+    sequence into a list that can be used with numpy dtype load data from the
     file with np.frombuffer().
 
     Some column IDs refer to small values which are packed into a single byte.
     The second return value is a dict describing the bit masks with which to
     extract these columns from the flags byte.
 
+    If error_on_unknown_column is True, an error will be raised if an unknown
+    column ID is encountered. If it is False, a warning will be emited and attempts
+    will be made to read the column with a few different dtypes.
+
+
     """
     type_list = []
     field_name_counts = defaultdict(int)
@@ -469,11 +477,19 @@ def VMPdata_dtype_from_colIDs(colIDs):
                 unique_field_name = field_name
             type_list.append((unique_field_name, field_type))
         else:
-            raise NotImplementedError(
-                "Column ID {cid} after column {prev} "
-                "is unknown".format(cid=colID, prev=type_list[-1][0])
+            if error_on_unknown_column:
+                raise NotImplementedError(
+                    "Column ID {cid} after column {prev} is unknown".format(
+                        cid=colID, prev=type_list[-1][0]
+                    )
+                )
+            warnings.warn(
+                "Unknown column ID %d -- will attempt to read as common dtypes"
+                % colID
             )
-    return np.dtype(type_list), flags_dict
+            type_list.append(("unknown_colID_%d" % colID, UNKNOWN_COLUMN_TYPE_HIERARCHY[0]))
+
+    return type_list, flags_dict
 
 
 def read_VMP_modules(fileobj, read_module_data=True):
@@ -544,7 +560,17 @@ class MPRfile:
     enddate - The date when the experiment finished
     """
 
-    def __init__(self, file_or_path):
+    def __init__(self, file_or_path, error_on_unknown_column: bool = True):
+        """Pass an EC-lab .mpr file to be parsed.
+
+        Parameters:
+            file_or_path: Either the open file data or a path to it.
+            error_on_unknown_column: Whether or not to raise an error if an
+                unknown column ID is encountered. A warning will be emited and
+                the column will be added 'unknown_<colID>', with an attempt to read
+                it with a few different dtypes.
+
+        """
         self.loop_index = None
         if isinstance(file_or_path, str):
             mpr_file = open(file_or_path, "rb")
@@ -596,8 +622,41 @@ class MPRfile:
 
         assert not any(remaining_headers)
 
-        self.dtype, self.flags_dict = VMPdata_dtype_from_colIDs(column_types)
-        self.data = np.frombuffer(main_data, dtype=self.dtype)
+        dtypes, self.flags_dict = VMPdata_dtype_from_colIDs(
+            column_types, error_on_unknown_column=error_on_unknown_column
+        )
+
+        unknown_cols = []
+        # Iteratively work through the unknown columns and try to read them
+        if not error_on_unknown_column:
+            for col, _ in dtypes:
+                if col.startswith("unknown_colID"):
+                    unknown_cols.append(col)
+
+        if unknown_cols:
+            # create a list of all possible combinations of dtypes
+            # for the unknown columns
+            from itertools import product
+            perms = product(UNKNOWN_COLUMN_TYPE_HIERARCHY, repeat=len(unknown_cols))
+            for perm in perms:
+                for unknown_col_ind, c in enumerate(unknown_cols):
+                    for ind, (col, _) in enumerate(dtypes):
+                        if c == col:
+                            dtypes[ind] = (col, perm[unknown_col_ind])
+
+                try:
+                    self.dtype = np.dtype(dtypes)
+                    self.data = np.frombuffer(main_data, dtype=self.dtype)
+                    break
+                except ValueError:
+                    continue
+            else:
+                raise RuntimeError("Unable to read data for unknown columns %s with any of the common dtypes %s", unknown_cols, UNKNOWN_COLUMN_TYPE_HIERARCHY)
+
+        else:
+            self.dtype = np.dtype(dtypes)
+            self.data = np.frombuffer(main_data, dtype=self.dtype)
+
         assert self.data.shape[0] == n_data_points
 
         # No idea what these 'column types' mean or even if they are actually