option for deterministic descriptions of sets & tables

src/Desc.cc

@@ -40,6 +40,7 @@ ODesc::ODesc(DescType t, File* arg_f)
 	indent_level = 0;
 	is_short = false;
 	want_quotes = false;
+	want_determinism = false;
 	do_flush = true;
 	include_stats = false;
 	indent_with_spaces = 0;

src/Desc.h

@@ -47,6 +47,11 @@ public:
 	bool WantQuotes() const	{ return want_quotes; }
 	void SetQuotes(bool q)	{ want_quotes = q; }
 
+	// Whether to ensure deterministic output (for example, when
+	// describing TableVal's).
+	bool WantDeterminism() const	{ return want_determinism; }
+	void SetDeterminism(bool d)	{ want_determinism = d; }
+
 	// Whether we want to print statistics like access time and execution
 	// count where available.
 	bool IncludeStats() const	{ return include_stats; }
@@ -122,7 +127,7 @@ public:
 				Add("\n", 0);
 			}
 
-	// Bypasses the escaping enabled via SetEscape().
+	// Bypasses the escaping enabled via EnableEscaping().
 	void AddRaw(const char* s, int len)	{ AddBytesRaw(s, len); }
 	void AddRaw(const std::string &s)		{ AddBytesRaw(s.data(), s.size()); }
 
@@ -194,6 +199,7 @@ protected:
 	bool escape;	// escape unprintable characters in output?
 	bool is_short;
 	bool want_quotes;
+	bool want_determinism;
 
 	int indent_with_spaces;
 

src/Val.cc

@@ -2334,6 +2334,9 @@ void TableVal::Describe(ODesc* d) const
 		d->PushIndent();
 		}
 
+	bool determ = d->WantDeterminism();
+	std::vector<std::string> elem_descs;
+
 	auto iter = table_val->begin();
 
 	for ( int i = 0; i < n; ++i )
@@ -2347,7 +2350,10 @@ void TableVal::Describe(ODesc* d) const
 		auto vl = table_hash->RecoverVals(*k);
 		int dim = vl->Length();
 
-		if ( i > 0 )
+		ODesc intermediary_d;
+		ODesc* d_ptr = determ ? &intermediary_d : d;
+
+		if ( ! determ && i > 0 )
 			{
 			if ( ! d->IsBinary() )
 				d->Add(",");
@@ -2358,42 +2364,67 @@ void TableVal::Describe(ODesc* d) const
 		if ( d->IsReadable() )
 			{
 			if ( dim != 1 || ! table_type->IsSet() )
-				d->Add("[");
+				d_ptr->Add("[");
 			}
 		else
 			{
-			d->Add(dim);
-			d->SP();
+			d_ptr->Add(dim);
+			d_ptr->SP();
 			}
 
-		vl->Describe(d);
+		vl->Describe(d_ptr);
 
 		if ( table_type->IsSet() )
 			{ // We're a set, not a table.
 			if ( d->IsReadable() )
 				if ( dim != 1 )
-					d->AddSP("]");
+					d_ptr->AddSP("]");
 			}
 		else
 			{
 			if ( d->IsReadable() )
-				d->AddSP("] =");
+				d_ptr->AddSP("] =");
 			if ( v->GetVal() )
-				v->GetVal()->Describe(d);
+				v->GetVal()->Describe(d_ptr);
 			}
 
 		if ( d->IsReadable() && ! d->IsShort() && d->IncludeStats() )
 			{
-			d->Add(" @");
-			d->Add(util::detail::fmt_access_time(v->ExpireAccessTime()));
+			d_ptr->Add(" @");
+			d_ptr->Add(util::detail::fmt_access_time(v->ExpireAccessTime()));
 			}
 
+		if ( determ )
+			elem_descs.emplace_back(d_ptr->Description());
+
 		++iter;
 		}
 
 	if ( iter != table_val->end() )
 		reporter->InternalError("hash table overflow in TableVal::Describe");
 
+	if ( determ )
+		{
+		sort(elem_descs.begin(), elem_descs.end());
+
+		iter = table_val->begin();
+		bool did_elems = false;
+
+		for ( const auto& ed : elem_descs )
+			{
+			if ( did_elems )
+				{
+				if ( ! d->IsBinary() )
+					d->Add(",");
+
+				d->NL();
+				}
+
+			d->Add(ed);
+			did_elems = true;
+			}
+		}
+
 	if ( d->IsPortable() || d->IsReadable() )
 		{
 		d->PopIndent();
@@ -3501,7 +3532,8 @@ void describe_vals(const std::vector<ValPtr>& vals,
 		if ( i > offset && d->IsReadable() && d->Style() != RAW_STYLE )
 			d->Add(", ");
 
-		vals[i]->Describe(d);
+		if ( vals[i] )
+			vals[i]->Describe(d);
 		}
 	}