Fix two problems: 1. The date strings in Times were completely wrong (because...

Fix two problems: 1. The date strings in Times were completely wrong (because it was assumed the values in MJD were MATLAB dates (not Modified Julian Days). Fixing that problem highlighted some possible limitations in mjulian2greg in which some second values are output as 60 (which is obviously wrong). So, some additional checks on the values of the seconds are performed, and times adjusted as necessary. 2. The Code to get the list of open boundary node IDs was a bit ugly, so I rewrote it to be a bit prettier. The old version is left in as a comment so that it can be reverted easily if necessary

fvcom_prepro/write_FVCOM_elevtide.m

@@ -38,15 +38,23 @@ subname = 'write_FVCOM_elevtide';
 if(report); fprintf('\n'); end
 if(report); fprintf(['begin : ' subname '\n']); end
 
-cnt = 0;
-ObcNodes = nan(1,sum(Mobj.nObcNodes));
-for ob=1:Mobj.nObs
-	nObcs = Mobj.nObcNodes(ob);
-	for j=1:nObcs
-		cnt = cnt + 1;
-		ObcNodes(cnt) = Mobj.obc_nodes(ob,j);  % set open boundary nodes
-    end
-end
+% Get a list of the open boundary nodes. Transpose Mobj.obc_nodes so the
+% order of the boundary nodes is preserved.
+tmpObcNodes = Mobj.obc_nodes';
+% Flip it back so it's the same shape as it would have been using the old
+% code (see below for the old way).
+ObcNodes = tmpObcNodes(tmpObcNodes~=0)';
+
+% cnt = 0;
+% ObcNodes = nan(1,sum(Mobj.nObcNodes));
+% for ob=1:Mobj.nObs
+% 	nObcs = Mobj.nObcNodes(ob);
+% 	for j=1:nObcs
+% 		cnt = cnt + 1;
+% 		ObcNodes(cnt) = Mobj.obc_nodes(ob,j);  % set open boundary nodes
+%     end
+% end
+
 
 %------------------------------------------------------------------------------
 % Sanity check on input and dimensions
@@ -124,7 +132,21 @@ netcdf.putVar(nc,itime_varid,floor(MJD));
 netcdf.putVar(nc,itime2_varid,0,nTimes,mod(MJD,1)*24*3600*1000);
 nStringOut = char();
 for i=1:nTimes
-    nStringOut = [nStringOut, sprintf('%04i/%02i/%02i %02i:%02i:%02i       ',datevec(MJD(i)))];
+    [nYr, nMon, nDay, nHour, nMin, nSec] = mjulian2greg(MJD(i));
+    if strcmp(sprintf('%02i', nSec), '60')
+        % Fix some weirdness with mjulian2greg. I think this is caused by
+        % rounding errors. My testing suggests this is not a problem around
+        % midnight, so the number of days (and thus possibly months and
+        % years) is unaffected.
+        if mod(nMin + 1, 60) == 0
+            % Up the hour by one too
+            nHour = mod(nHour + 1, 24);
+        end
+        nMin = mod(nMin + 1, 60);
+        nSec = 0;
+    end
+    nDate = [nYr, nMon, nDay, nHour, nMin, nSec];
+    nStringOut = [nStringOut, sprintf('%04i/%02i/%02i %02i:%02i:%02i       ',nDate)];
 end
 netcdf.putVar(nc,Times_varid,nStringOut);
 netcdf.putVar(nc,elevation_varid,Mobj.surfaceElevation);