From c23464c79562aafb01c8c435665372a03bb85f0d Mon Sep 17 00:00:00 2001
From: Florez Ospina Juan Felipe <juan.florez-ospina@psi.ch>
Date: Tue, 15 Aug 2023 13:25:15 +0200
Subject: [PATCH] Modified code to write dataset and attributes' names with the
 indices of the original table's columns. The indices can then inferred by
 examining the names.

---
 import/save_table_as_hdf5.m | 67 +++++++++++++++++++++++++++++--------
 1 file changed, 53 insertions(+), 14 deletions(-)

diff --git a/import/save_table_as_hdf5.m b/import/save_table_as_hdf5.m
index f95cab5..7b6c1bf 100644
--- a/import/save_table_as_hdf5.m
+++ b/import/save_table_as_hdf5.m
@@ -16,12 +16,19 @@ else
     return
 end
 
+% Create array with indices of column_names
+column_names_idx = 1 : length(column_names);
+
 % Divide input_table into tables with numeric and cellstr variables/columns
 numeric_table = input_table(:,vartype('numeric')); % group attributes of a timestamped measurement
 cellstr_table = input_table(:,vartype('cellstr')); % group datasets of a timestamped measurement
 
+% Record columns/vars original index in reference table
 numeric_columns = numeric_table.Properties.VariableNames;
+numeric_columns_idx = argwhere_var_in_varset(numeric_columns,column_names);
+
 str_columns = cellstr_table.Properties.VariableNames;
+str_columns_idx = argwhere_var_in_varset(str_columns,column_names);
 
 % Based on homogeneous columns/cell arrays, we identify the remaining columns
 % with possibly heterogeneous entries. Ideally, these should store numeric
@@ -41,6 +48,8 @@ for i = 1 : length(input_table.Properties.VariableNames)
     end
 end
 
+rest_columns_idx = argwhere_var_in_varset(rest_columns,column_names);
+
 % Create an hdf5 file, where rows are seen as measurements/groups with
 % attributes given by numeric and str variables and datasets given by the 
 % remaining variables as long as they are numeric n-dimensional arrays.
@@ -51,13 +60,14 @@ cell_table = input_table(:,rest_columns);
 for i = 1 : rows
     group_name = strcat("/MEAS",string(i),"/");
     for j = 1 : length(rest_columns)
-        dataset_name = strcat("DS",string(j));
+        dataset_name = strcat("DS",string(rest_columns_idx(j)));
 
         % Some table entries do not provide direct access to the object
         % of interest, but may contain either {object} or {{object}}. 
         % The following loop solves that problem. TODO: is this generic enough? 
 
-        data = cell_table{i,rest_columns{j}}{1};%{1};
+        data = cell_table{i,column_names{rest_columns_idx(j)}}{1}; % <=> data = cell_table{i,rest_columns{j}}{1};
+
         while iscell(data)
             data = data{1};
         end
@@ -72,8 +82,13 @@ for i = 1 : rows
         h5create(filename, strcat(group_name,dataset_name),sz);
         % Dataset's attribute: Add attribute with dataset_name equal to column_name
         attr = "column_name";        
-        attr_value = rest_columns{j};
+        attr_value = column_names{rest_columns_idx(j)}; % <=> rest_columns{j};
         h5writeatt(filename,strcat(group_name,dataset_name),attr,attr_value)
+
+        %attr = "column_name_idx";        
+        %attr_value = rest_columns_idx(j); % <=> rest_columns{j};
+        %h5writeatt(filename,strcat(group_name,dataset_name),attr,attr_value)
+
         % Assign 'data' to the dataset. 'data' has to be of admissible type. 
         try            
             h5write(filename,strcat(group_name,dataset_name),data)
@@ -84,22 +99,46 @@ for i = 1 : rows
     end
 
     % Group's attribute: Assign group attributes
-    for attr_idx = 1 : max([length(numeric_columns),length(str_columns)])
+    for attr_idx = 1 : length(numeric_columns) 
+        attr = column_names{numeric_columns_idx(attr_idx)};% <=> attr = numeric_columns{attr_idx};
+        attr_value = numeric_table{i,attr};
 
-        if attr_idx <= length(numeric_columns)
-            attr = numeric_columns{attr_idx};
-            attr_value = numeric_table{i,numeric_columns{attr_idx}};
-            h5writeatt(filename,group_name,attr,attr_value)
-        end
+        column_idx_tag = ['ATTR',num2str(numeric_columns_idx(attr_idx)),'_'];
+        h5writeatt(filename,group_name,[column_idx_tag,attr],attr_value)
 
-        if attr_idx <= length(str_columns)
-            attr = str_columns{attr_idx};
-            attr_value = cellstr_table{i,str_columns{attr_idx}};
-            h5writeatt(filename,group_name,attr,attr_value)
-        end
+        %attr = "column_name_idx";        
+        %attr_value = numeric_columns_idx(attr_idx); % <=> rest_columns{j};
 
+        %h5writeatt(filename,group_name,attr,attr_value)
+    end
+
+    for attr_idx = 1 :  length(str_columns)
+    
+        attr = column_names{str_columns_idx(attr_idx)}; % <=> attr = str_columns{attr_idx}
+        attr_value = cellstr_table{i,attr};
+
+        column_idx_tag = ['ATTR',num2str(str_columns_idx(attr_idx)),'_'];
+        h5writeatt(filename,group_name,[column_idx_tag,attr],attr_value)
+
+        %attr = "column_name_idx";        
+        %attr_value = str_columns_idx(attr_idx); % <=> rest_columns{j};
+
+        %h5writeatt(filename,group_name,attr,attr_value)
     end
     %h5create(filename,"/g2/DS2",[20 Inf],"Chunksize",[5 5]);
 end
 
 success = 1;
+
+function idx = argwhere_var_in_varset(var,varset)
+    ref_idx = 1 : length(varset);
+    if iscellstr(var)
+        idx = [];
+        for i = 1 : length(var)
+            logical_idx = strcmp(var{i},varset); 
+            idx = [idx,ref_idx(logical_idx)];
+        end
+    else
+        logical_idx = strcmp(var,varset);    
+        idx = ref_idx(logical_idx);
+    end