public void AddPaths(DcColumn sd, DcColumn td, Mapping gMapping) // Add this pair by expanding it using the mapping
{
Debug.Assert(sd != null && sd.Input == SourceTab, "Wrong use: source path must start from the source table.");
Debug.Assert(td != null && td.Input == TargetTab, "Wrong use: target path must start from the target table.");
Debug.Assert(sd != null && sd.Output == gMapping.SourceTab, "Wrong use: source path must end where the mapping starts.");
Debug.Assert(td != null && td.Output == gMapping.TargetTab, "Wrong use: target path must end where the mapping starts.");
if (gMapping.Matches.Count == 0) // If there are no continuations then add only the starting segments (for example, for mappings between primitive tables)
{
ColumnPath sp = new ColumnPath(); // A path consists of one segment
sp.InsertLast(sd);
ColumnPath tp = new ColumnPath(); // A path consists of one segment
tp.InsertLast(td);
PathMatch match = new PathMatch(sp, tp);
Matches.Add(match);
}
foreach (PathMatch gMatch in gMapping.Matches)
{
ColumnPath sp = new ColumnPath(); // Create source path by concatenating one segment and continuation path from the mapping
sp.InsertLast(sd);
sp.InsertLast(gMatch.SourcePath);
ColumnPath tp = new ColumnPath(); // Create target path by concatenating one segment and continuation path from the mapping
tp.InsertLast(td);
tp.InsertLast(gMatch.TargetPath);
PathMatch match = new PathMatch(sp, tp);
Matches.Add(match);
}
}
public bool Compatible(PathMatch match) // The specified match is compatible with (does not contradicts to) this match
{
// Main rule: if a coverage condition holds for one path (in one or another direction) then it must hold for the other path (in the same direction)
// SourcePath -> TargetPath
if (match.SourcePath.StartsWith(SourcePath)) // The specified match continues this path
{
return(match.TargetPath.StartsWith(TargetPath)); // The same must be true for the other paths
}
else if (SourcePath.StartsWith(match.SourcePath)) // Opposite
{
return(TargetPath.StartsWith(match.TargetPath)); // The same must be true for the other paths
}
// TargetPath -> SourcePath
if (match.TargetPath.StartsWith(TargetPath))
{
return(match.SourcePath.StartsWith(SourcePath));
}
else if (TargetPath.StartsWith(match.TargetPath))
{
return(SourcePath.StartsWith(match.SourcePath));
}
// Neither source path nor target paths cover.
// This means they have some intersection (meet table they both continue).
// This meet point defines an implicit match which we do not check but it might contradict to other matches.
return(true);
}
public void AddMatches(List <PathMatch> matches) // Import all path matches (as new objects) that fit into this mapping
{
foreach (PathMatch m in matches)
{
PathMatch match = new PathMatch(m);
AddMatch(match);
}
}
public bool AreMatched(PathMatch match) // Determine if both paths are matched (true if there exist the same or more specific match)
{
foreach (PathMatch m in Matches)
{
if (m.Matches(match))
{
return(true);
}
}
return(false); // Not found
}
public bool Compatible(PathMatch match) // Determine if the specified match does not contradict to this mapping and can be added to it without removing existing matches
{
foreach (PathMatch m in Matches)
{
if (!m.Compatible(match))
{
return(false);
}
}
return(true); // Compatible and can be added
}
public void AddTargetToSchema(DcSchema schema = null) // Ensure that all target elements exist in the specified schema
{
// The mapping can reference new elements which are not in the schema yet so we try to find them and add if necessary
if (schema == null) // Find the schema from the mapping elements
{
PathMatch match = Matches.FirstOrDefault(m => m.TargetPath.Output.IsPrimitive);
schema = match != null ? match.TargetPath.Output.Schema : null; // We assume that primitive tables always have root defined (other tables might not have been added yet).
}
ColumnTree tree = GetTargetTree();
tree.AddToSchema(schema);
}
public virtual void FromJson(JObject json, DcSpace ws)
{
// No super-object
Similarity = (double)json["similarity"];
_sourceTab = (DcTable)Com.Schema.Utils.ResolveJsonRef((JObject)json["source_table"], ws);
_targetTab = (DcTable)Com.Schema.Utils.ResolveJsonRef((JObject)json["target_table"], ws);
// List of matches
foreach (JObject match in json["matches"])
{
PathMatch comMatch = (PathMatch)Com.Schema.Utils.CreateObjectFromJson(match);
if (comMatch != null)
{
comMatch.FromJson(match, ws);
this.Matches.Add(comMatch);
}
}
}
public void AddMatch(PathMatch match)
{
Debug.Assert(match != null && match.SourceTab == SourceTab, "Wrong use: source path must start from the source table.");
Debug.Assert(match != null && match.TargetTab == TargetTab, "Wrong use: target path must start from the target table.");
// In the case of coverage we do nothing. We want to keep more general matches (between tables) while more specific can be added/removed independently (if they satisfy them).
// ??? in the case of no coverage, they have a point of intersection. Then - ? Should this point of intersection be within the seconad path or otherwise constrained?
// Problem: two tables can be matched explcitly or implicitly, that is, their match logically follows from other path matches.
// Implicit match: intersections of two paths produce a more general match
// Alternative approach: Incrementally add path segments and checking consistency conditions
List <PathMatch> toRemove = new List <PathMatch>();
foreach (PathMatch m in Matches)
{
if (!m.Compatible(match))
{
toRemove.Add(m);
}
}
Matches.Add(match); // Now it is guaranteed to be compatible with all existing matches
}
/// <summary>
/// Create and initialize a new mapping which produces a flat target set with all primitive columns for copying primitive data from the source set.
/// Only identity (PK) source columns are expanded recursively.
/// For relational source, this means that all primitive columns of the source table will be mapped with their relational names, no FK-referenced tables will be joined and no artifical column names will be used.
/// If it is necessary to expand entity columns (non-PK columns of joined tables) then a different implementation is needed (which will require joins, artifical column/path names etc.)
/// </summary>
public Mapping CreatePrimitive(DcTable sourceSet, DcTable targetSet, DcSchema targetSchema)
{
Debug.Assert(!sourceSet.IsPrimitive && !targetSet.IsPrimitive, "Wrong use: copy mapping can be created for only non-primitive tables.");
Debug.Assert(targetSchema != null || targetSet.Schema != null, "Wrong use: target schema must be specified.");
Mapping map = new Mapping(sourceSet, targetSet);
DcSchema sourceSchema = map.SourceTab.Schema;
if (targetSchema == null)
{
targetSchema = targetSet.Schema;
}
ColumnPath sp;
ColumnPath tp;
DcColumn td;
PathMatch match;
if (sourceSchema is SchemaOledb)
{
TableRel set = (TableRel)map.SourceTab;
foreach (ColumnAtt att in set.GreaterPaths)
{
sp = new ColumnAtt(att);
// Recommend matching target type (mapping primitive types)
this.MapPrimitiveSet(att.Output, targetSchema);
DcTable targetType = this.GetBestTargetSet(att.Output, targetSchema);
td = new Schema.Column(att.RelationalColumnName, map.TargetTab, targetType, att.IsKey, false);
tp = new ColumnPath(td);
tp.Name = sp.Name;
match = new PathMatch(sp, tp, 1.0);
map.Matches.Add(match);
}
}
else if (sourceSchema is SchemaCsv)
{
DcTable set = (DcTable)map.SourceTab;
foreach (DcColumn sd in set.Columns)
{
if (sd.IsSuper)
{
continue;
}
// Recommend matching target type (mapping primitive types)
//this.MapPrimitiveSet(sd, targetSchema);
//ComTable targetType = this.GetBestTargetSet(sd.Output, targetSchema);
//
// Analyze sample values of sd and choose the most specific target type
//
List <string> values = ((ColumnCsv)sd).SampleValues;
string targetTypeName;
if (Com.Schema.Utils.isInt32(values.ToArray()))
{
targetTypeName = "Integer";
}
else if (Com.Schema.Utils.isDouble(values.ToArray()))
{
targetTypeName = "Double";
}
else
{
targetTypeName = "String";
}
DcTable targetType = targetSchema.GetPrimitiveType(targetTypeName);
td = targetSchema.Space.CreateColumn(sd.Name, map.TargetTab, targetType, sd.IsKey);
sp = new ColumnPath(sd);
tp = new ColumnPath(td);
match = new PathMatch(sp, tp, 1.0);
map.Matches.Add(match);
}
}
return(map);
}
public PathMatch(PathMatch m)
{
SourcePath = new ColumnPath(m.SourcePath);
TargetPath = new ColumnPath(m.TargetPath);
Similarity = m.Similarity;
}
public bool Matches(PathMatch match) // This is more specific (longer) than argument
{
return(MatchesSource(match.SourcePath) && MatchesTarget(match.TargetPath));
}
