Remove importExcelModel and convert tutorial models to YAML (#631)

* Remove importExcelModel and convert tutorial models to YAML

- Remove io/importExcelModel.m and its generated documentation
- Convert tutorial models small, smallYeast and smallYeastBad from Excel
  to YAML; rewrite tutorials 1-4 and their solutions to load models via
  importModel/readYAMLmodel instead of importExcelModel
- Drop the smallYeastBad2 section of tutorial 4 (duplicate metabolite
  names cannot be represented in SBML/YAML)
- readYAMLmodel: keep rxnGeneMat dimensionally consistent for gene-less
  models, which were left with an empty matrix that broke removeReactions
- followChanged: handle the case where no reactions differ, where the
  empty reaction list was expanded to all reactions by constructEquations
- fillGaps: guard against an empty toCheck list, which haveFlux otherwise
  interprets as all reactions
- tutorial3_solutions: set upper bounds before lower bounds so a lower
  bound never transiently exceeds the previous upper bound
- tutorial4: define the reference model before it is used
- Remove Excel import coverage from tIO and importExportTests

* Remove orphaned Excel model files from tutorial

With importExcelModel gone, nothing can read the Excel source models, so
the now-unreferenced tutorial/{small,smallYeast,smallYeastBad,
smallYeastBad2}.xlsx and "iAL1006 v1.00.xlsx" are removed. The small,
smallYeast and smallYeastBad models remain as YAML; iAL1006 remains as
SBML. The empty.{xml,xlsx,yml,mat} set is kept on purpose, as each format
is used to test a different importer (importModel, loadWorkbook/loadSheet,
readYAMLmodel/parseYAML and the MAT-based importExportTests). Update the
tutorial1 header that referred to the removed Excel file.

Author: edkerk

analysis/followChanged.m

@@ -100,7 +100,13 @@ function followChanged(model,fluxesA,fluxesB, varargin)
 zeroFluxes=ineither(find(fluxesA(ineither)==0));
 fluxIndexes=unique([fluxIndexes zeroFluxes(find(abs(fluxesB(zeroFluxes))>=cutOffFlux))']);
 
-formulas=constructEquations(model,model.rxns(fluxIndexes));
+if isempty(fluxIndexes)
+    % No reactions changed; avoid constructEquations interpreting an empty
+    % reaction list as "all reactions"
+    formulas={};
+else
+    formulas=constructEquations(model,model.rxns(fluxIndexes));
+end
 
 if cutOffFluxSupplied
     if cutOffDiffSupplied

doc/io/importExcelModel.html

@@ -189,12 +189,17 @@
     %merged model
     toCheck=intersect(allModels.rxns(strcmp(allModels.rxnFrom,model.id)),model.rxns(~originalFlux));
 
-    %Get the ones that still cannot carry flux
-    I=haveFlux(allModels,1,toCheck);
-    
-    %Get the reactions that can't carry flux in the original model, but can
-    %in the merged one
-    K=toCheck(I);
+    %Get the ones that still cannot carry flux. Guard against an empty
+    %toCheck, since haveFlux interprets an empty reaction list as "all
+    %reactions" rather than "none"
+    if isempty(toCheck)
+        K=toCheck;
+    else
+        I=haveFlux(allModels,1,toCheck);
+        %Get the reactions that can't carry flux in the original model, but can
+        %in the merged one
+        K=toCheck(I);
+    end
 
     %This is a temporary thing to only look at the non-reversible rxns.
     %This is because all reversible rxns can have a flux in the