void DiagList::Copy(const DiagList &DL)
{
Clear();
unsigned uCount = DL.GetCount();
for (unsigned i = 0; i < uCount; ++i)
Add(DL.Get(i));
}
void DiagListToDPRegionList(const DiagList &DL, DPRegionList &RL,
unsigned uLengthA, unsigned uLengthB)
{
if (g_uDiagMargin > g_uMinDiagLength/2)
Quit("Invalid parameters, diagmargin=%d must be <= 2*diaglength=%d",
g_uDiagMargin, g_uMinDiagLength);
unsigned uStartPosA = 0;
unsigned uStartPosB = 0;
const unsigned uDiagCount = DL.GetCount();
DPRegion r;
for (unsigned uDiagIndex = 0; uDiagIndex < uDiagCount; ++uDiagIndex)
{
const Diag &d = DL.Get(uDiagIndex);
assert(d.m_uLength >= g_uMinDiagLength);
const unsigned uStartVertexA = d.m_uStartPosA + g_uDiagMargin - 1;
const unsigned uStartVertexB = d.m_uStartPosB + g_uDiagMargin - 1;
const unsigned uEndVertexA = d.m_uStartPosA + d.m_uLength - g_uDiagMargin;
const unsigned uEndVertexB = d.m_uStartPosB + d.m_uLength - g_uDiagMargin;
r.m_Type = DPREGIONTYPE_Rect;
r.m_Rect.m_uStartPosA = uStartPosA;
r.m_Rect.m_uStartPosB = uStartPosB;
assert(uStartVertexA + 1 >= uStartPosA);
assert(uStartVertexB + 1 >= uStartPosB);
r.m_Rect.m_uLengthA = uStartVertexA + 1 - uStartPosA;
r.m_Rect.m_uLengthB = uStartVertexB + 1 - uStartPosB;
RL.Add(r);
if (uEndVertexA > uStartVertexA + 1)
{
const unsigned uDiagLengthMinusCaps = uEndVertexA - uStartVertexA - 1;
r.m_Type = DPREGIONTYPE_Diag;
r.m_Diag.m_uStartPosA = uStartVertexA + 1;
r.m_Diag.m_uStartPosB = uStartVertexB + 1;
assert(uEndVertexA - uStartVertexA == uEndVertexB - uStartVertexB);
r.m_Diag.m_uLength = uEndVertexA - uStartVertexA - 1;
RL.Add(r);
}
uStartPosA = uEndVertexA;
uStartPosB = uEndVertexB;
}
assert((int) uLengthA - (int) uStartPosA >= (int) g_uDiagMargin);
assert((int) uLengthB - (int) uStartPosB >= (int) g_uDiagMargin);
r.m_Type = DPREGIONTYPE_Rect;
r.m_Rect.m_uStartPosA = uStartPosA;
r.m_Rect.m_uStartPosB = uStartPosB;
assert(uLengthA >= uStartPosA);
assert(uLengthB >= uStartPosB);
r.m_Rect.m_uLengthA = uLengthA - uStartPosA;
r.m_Rect.m_uLengthB = uLengthB - uStartPosB;
RL.Add(r);
}
/// CompareDiagLists - Compare two diagnostic lists and return the difference
/// between them.
///
static bool CompareDiagLists(SourceManager &SourceMgr,
const_diag_iterator d1_begin,
const_diag_iterator d1_end,
const_diag_iterator d2_begin,
const_diag_iterator d2_end,
const char *MsgLeftOnly,
const char *MsgRightOnly) {
DiagList LeftOnly;
DiagList Left(d1_begin, d1_end);
DiagList Right(d2_begin, d2_end);
for (const_diag_iterator I = Left.begin(), E = Left.end(); I != E; ++I) {
unsigned LineNo1 = SourceMgr.getInstantiationLineNumber(I->first);
const std::string &Diag1 = I->second;
DiagList::iterator II, IE;
for (II = Right.begin(), IE = Right.end(); II != IE; ++II) {
unsigned LineNo2 = SourceMgr.getInstantiationLineNumber(II->first);
if (LineNo1 != LineNo2) continue;
const std::string &Diag2 = II->second;
if (Diag2.find(Diag1) != std::string::npos ||
Diag1.find(Diag2) != std::string::npos) {
break;
}
}
if (II == IE) {
// Not found.
LeftOnly.push_back(*I);
} else {
// Found. The same cannot be found twice.
Right.erase(II);
}
}
// Now all that's left in Right are those that were not matched.
return PrintProblem(SourceMgr, LeftOnly.begin(), LeftOnly.end(), MsgLeftOnly)
| PrintProblem(SourceMgr, Right.begin(), Right.end(), MsgRightOnly);
}
// Merge diagonals that are continuations of each other with
// short breaks of up to length g_uMaxDiagBreak.
// In a sorted list of diagonals, we only have to check
// consecutive entries.
void MergeDiags(DiagList &DL)
{
unsigned &g_uMaxDiagBreak = getMuscleContext()->params.g_uMaxDiagBreak;
return;
#if DEBUG
if (!DL.IsSorted())
Quit("MergeDiags: !IsSorted");
#endif
// TODO: Fix this!
// Breaks must be with no offset (no gaps)
const unsigned uCount = DL.GetCount();
if (uCount <= 1)
return;
DiagList NewList;
Diag MergedDiag;
const Diag *ptrPrev = &DL.Get(0);
for (unsigned i = 1; i < uCount; ++i)
{
const Diag *ptrDiag = &DL.Get(i);
unsigned uBreakLength = DiagBreak(*ptrPrev, *ptrDiag);
if (uBreakLength <= g_uMaxDiagBreak)
{
MergedDiag.m_uStartPosA = ptrPrev->m_uStartPosA;
MergedDiag.m_uStartPosB = ptrPrev->m_uStartPosB;
MergedDiag.m_uLength = ptrPrev->m_uLength + ptrDiag->m_uLength
+ uBreakLength;
ptrPrev = &MergedDiag;
}
else
{
NewList.Add(*ptrPrev);
ptrPrev = ptrDiag;
}
}
NewList.Add(*ptrPrev);
DL.Copy(NewList);
}
/// CheckResults - This compares the expected results to those that
/// were actually reported. It emits any discrepencies. Return "true" if there
/// were problems. Return "false" otherwise.
///
static bool CheckResults(Preprocessor &PP,
const DiagList &ExpectedErrors,
const DiagList &ExpectedWarnings,
const DiagList &ExpectedNotes) {
const DiagnosticClient *DiagClient = PP.getDiagnostics().getClient();
assert(DiagClient != 0 &&
"DiagChecker requires a valid TextDiagnosticBuffer");
const TextDiagnosticBuffer &Diags =
static_cast<const TextDiagnosticBuffer&>(*DiagClient);
SourceManager &SourceMgr = PP.getSourceManager();
// We want to capture the delta between what was expected and what was
// seen.
//
// Expected \ Seen - set expected but not seen
// Seen \ Expected - set seen but not expected
bool HadProblem = false;
// See if there are error mismatches.
HadProblem |= CompareDiagLists(SourceMgr,
ExpectedErrors.begin(), ExpectedErrors.end(),
Diags.err_begin(), Diags.err_end(),
"Errors expected but not seen:",
"Errors seen but not expected:");
// See if there are warning mismatches.
HadProblem |= CompareDiagLists(SourceMgr,
ExpectedWarnings.begin(),
ExpectedWarnings.end(),
Diags.warn_begin(), Diags.warn_end(),
"Warnings expected but not seen:",
"Warnings seen but not expected:");
// See if there are note mismatches.
HadProblem |= CompareDiagLists(SourceMgr,
ExpectedNotes.begin(),
ExpectedNotes.end(),
Diags.note_begin(), Diags.note_end(),
"Notes expected but not seen:",
"Notes seen but not expected:");
return HadProblem;
}
void FindDiagsNuc(const ProfPos *PX, unsigned uLengthX, const ProfPos *PY,
unsigned uLengthY, DiagList &DL)
{
MuscleContext *ctx = getMuscleContext();
ALPHA &g_Alpha = ctx->alpha.g_Alpha;
unsigned* TuplePos = ctx->finddiagsn.TuplePos;
unsigned &g_uMinDiagLength = ctx->params.g_uMinDiagLength;
if (ALPHA_DNA != g_Alpha && ALPHA_RNA != g_Alpha)
Quit("FindDiagsNuc: requires nucleo alphabet");
DL.Clear();
// 16 is arbitrary slop, no principled reason for this.
if (uLengthX < K + 16 || uLengthY < K + 16)
return;
// Set A to shorter profile, B to longer
const ProfPos *PA;
const ProfPos *PB;
unsigned uLengthA;
unsigned uLengthB;
bool bSwap;
if (uLengthX < uLengthY)
{
bSwap = false;
PA = PX;
PB = PY;
uLengthA = uLengthX;
uLengthB = uLengthY;
}
else
{
bSwap = true;
PA = PY;
PB = PX;
uLengthA = uLengthY;
uLengthB = uLengthX;
}
#if TRACE
Log("FindDiagsNuc(LengthA=%d LengthB=%d\n", uLengthA, uLengthB);
#endif
// Build tuple map for the longer profile, B
if (uLengthB < K)
Quit("FindDiags: profile too short");
memset(TuplePos, EMPTY, sizeof(TuplePos));
for (unsigned uPos = 0; uPos < uLengthB - K; ++uPos)
{
const unsigned uTuple = GetTuple(PB, uPos);
if (EMPTY == uTuple)
continue;
TuplePos[uTuple] = uPos;
}
// Find matches
for (unsigned uPosA = 0; uPosA < uLengthA - K; ++uPosA)
{
const unsigned uTuple = GetTuple(PA, uPosA);
if (EMPTY == uTuple)
continue;
const unsigned uPosB = TuplePos[uTuple];
if (EMPTY == uPosB)
continue;
// This tuple is found in both profiles
unsigned uStartPosA = uPosA;
unsigned uStartPosB = uPosB;
// Try to extend the match forwards
unsigned uEndPosA = uPosA + K - 1;
unsigned uEndPosB = uPosB + K - 1;
for (;;)
{
if (uLengthA - 1 == uEndPosA || uLengthB - 1 == uEndPosB)
break;
const unsigned uAAGroupA = PA[uEndPosA+1].m_uResidueGroup;
if (RESIDUE_GROUP_MULTIPLE == uAAGroupA)
break;
const unsigned uAAGroupB = PB[uEndPosB+1].m_uResidueGroup;
if (RESIDUE_GROUP_MULTIPLE == uAAGroupB)
break;
if (uAAGroupA != uAAGroupB)
break;
++uEndPosA;
++uEndPosB;
}
uPosA = uEndPosA;
#if TRACE
{
Log("Match: A %4u-%4u ", uStartPosA, uEndPosA);
for (unsigned n = uStartPosA; n <= uEndPosA; ++n)
Log("%c", LetterToChar(PA[n].m_uResidueGroup));
Log("\n");
Log(" B %4u-%4u ", uStartPosB, uEndPosB);
for (unsigned n = uStartPosB; n <= uEndPosB; ++n)
Log("%c", LetterToChar(PB[n].m_uResidueGroup));
Log("\n");
}
#endif
const unsigned uLength = uEndPosA - uStartPosA + 1;
assert(uEndPosB - uStartPosB + 1 == uLength);
if (uLength >= g_uMinDiagLength)
{
if (bSwap)
DL.Add(uStartPosB, uStartPosA, uLength);
else
DL.Add(uStartPosA, uStartPosB, uLength);
}
}
}
void FindDiags(const ProfPos *PX, unsigned uLengthX, const ProfPos *PY,
unsigned uLengthY, DiagList &DL)
{
if (ALPHA_Amino != g_Alpha)
Quit("FindDiags: requires amino acid alphabet");
DL.Clear();
if (uLengthX < 12 || uLengthY < 12)
return;
// Set A to shorter profile, B to longer
const ProfPos *PA;
const ProfPos *PB;
unsigned uLengthA;
unsigned uLengthB;
bool bSwap;
if (uLengthX < uLengthY)
{
bSwap = false;
PA = PX;
PB = PY;
uLengthA = uLengthX;
uLengthB = uLengthY;
}
else
{
bSwap = true;
PA = PY;
PB = PX;
uLengthA = uLengthY;
uLengthB = uLengthX;
}
// Build tuple map for the longer profile, B
if (uLengthB < KTUP)
Quit("FindDiags: profile too short");
memset(TuplePos, EMPTY, sizeof(TuplePos));
for (unsigned uPos = 0; uPos < uLengthB - KTUP; ++uPos)
{
const unsigned uTuple = GetTuple(PB, uPos);
if (EMPTY == uTuple)
continue;
TuplePos[uTuple] = uPos;
}
// Find matches
for (unsigned uPosA = 0; uPosA < uLengthA - KTUP; ++uPosA)
{
const unsigned uTuple = GetTuple(PA, uPosA);
if (EMPTY == uTuple)
continue;
const unsigned uPosB = TuplePos[uTuple];
if (EMPTY == uPosB)
continue;
// This tuple is found in both profiles
unsigned uStartPosA = uPosA;
unsigned uStartPosB = uPosB;
// Try to extend the match forwards
unsigned uEndPosA = uPosA + KTUP - 1;
unsigned uEndPosB = uPosB + KTUP - 1;
for (;;)
{
if (uLengthA - 1 == uEndPosA || uLengthB - 1 == uEndPosB)
break;
const unsigned uAAGroupA = PA[uEndPosA+1].m_uResidueGroup;
if (RESIDUE_GROUP_MULTIPLE == uAAGroupA)
break;
const unsigned uAAGroupB = PB[uEndPosB+1].m_uResidueGroup;
if (RESIDUE_GROUP_MULTIPLE == uAAGroupB)
break;
if (uAAGroupA != uAAGroupB)
break;
++uEndPosA;
++uEndPosB;
}
uPosA = uEndPosA;
#if TRACE
{
Log("Match: A %4u-%4u ", uStartPosA, uEndPosA);
for (unsigned n = uStartPosA; n <= uEndPosA; ++n)
Log("%c", 'A' + PA[n].m_uResidueGroup);
Log("\n");
Log(" B %4u-%4u ", uStartPosB, uEndPosB);
for (unsigned n = uStartPosB; n <= uEndPosB; ++n)
Log("%c", 'A' + PB[n].m_uResidueGroup);
Log("\n");
}
#endif
const unsigned uLength = uEndPosA - uStartPosA + 1;
assert(uEndPosB - uStartPosB + 1 == uLength);
if (uLength >= g_uMinDiagLength)
{
if (bSwap)
DL.Add(uStartPosB, uStartPosA, uLength);
else
DL.Add(uStartPosA, uStartPosB, uLength);
}
}
}
/// FindDiagnostics - Go through the comment and see if it indicates expected
/// diagnostics. If so, then put them in a diagnostic list.
///
static void FindDiagnostics(const char *CommentStart, unsigned CommentLen,
DiagList &ExpectedDiags,
Preprocessor &PP, SourceLocation Pos,
const char *ExpectedStr) {
const char *CommentEnd = CommentStart+CommentLen;
unsigned ExpectedStrLen = strlen(ExpectedStr);
// Find all expected-foo diagnostics in the string and add them to
// ExpectedDiags.
while (CommentStart != CommentEnd) {
CommentStart = std::find(CommentStart, CommentEnd, 'e');
if (unsigned(CommentEnd-CommentStart) < ExpectedStrLen) return;
// If this isn't expected-foo, ignore it.
if (memcmp(CommentStart, ExpectedStr, ExpectedStrLen)) {
++CommentStart;
continue;
}
CommentStart += ExpectedStrLen;
// Skip whitespace.
while (CommentStart != CommentEnd &&
isspace(CommentStart[0]))
++CommentStart;
// Default, if we find the '{' now, is 1 time.
int Times = 1;
int Temp = 0;
// In extended syntax, there could be a digit now.
while (CommentStart != CommentEnd &&
CommentStart[0] >= '0' && CommentStart[0] <= '9') {
Temp *= 10;
Temp += CommentStart[0] - '0';
++CommentStart;
}
if (Temp > 0)
Times = Temp;
// Skip whitespace again.
while (CommentStart != CommentEnd &&
isspace(CommentStart[0]))
++CommentStart;
// We should have a {{ now.
if (CommentEnd-CommentStart < 2 ||
CommentStart[0] != '{' || CommentStart[1] != '{') {
if (std::find(CommentStart, CommentEnd, '{') != CommentEnd)
EmitError(PP, Pos, "bogus characters before '{{' in expected string");
else
EmitError(PP, Pos, "cannot find start ('{{') of expected string");
return;
}
CommentStart += 2;
// Find the }}.
const char *ExpectedEnd = CommentStart;
while (1) {
ExpectedEnd = std::find(ExpectedEnd, CommentEnd, '}');
if (CommentEnd-ExpectedEnd < 2) {
EmitError(PP, Pos, "cannot find end ('}}') of expected string");
return;
}
if (ExpectedEnd[1] == '}')
break;
++ExpectedEnd; // Skip over singular }'s
}
std::string Msg(CommentStart, ExpectedEnd);
std::string::size_type FindPos;
while ((FindPos = Msg.find("\\n")) != std::string::npos)
Msg.replace(FindPos, 2, "\n");
// Add is possibly multiple times.
for (int i = 0; i < Times; ++i)
ExpectedDiags.push_back(std::make_pair(Pos, Msg));
CommentStart = ExpectedEnd;
}
}
SCORE GlobalAlignDiags(const ProfPos *PA, unsigned uLengthA, const ProfPos *PB,
unsigned uLengthB, PWPath &Path)
{
#if LIST_DIAGS
TICKS t1 = GetClockTicks();
#endif
DiagList DL;
if (ALPHA_Amino == g_Alpha)
FindDiags(PA, uLengthA, PB, uLengthB, DL);
else if (ALPHA_DNA == g_Alpha || ALPHA_RNA == g_Alpha)
FindDiagsNuc(PA, uLengthA, PB, uLengthB, DL);
else
Quit("GlobalAlignDiags: bad alpha");
#if TRACE
Log("GlobalAlignDiags, diag list:\n");
DL.LogMe();
#endif
DL.Sort();
DL.DeleteIncompatible();
#if TRACE
Log("After DeleteIncompatible:\n");
DL.LogMe();
#endif
MergeDiags(DL);
#if TRACE
Log("After MergeDiags:\n");
DL.LogMe();
#endif
DPRegionList RL;
DiagListToDPRegionList(DL, RL, uLengthA, uLengthB);
#if TRACE
Log("RegionList:\n");
RL.LogMe();
#endif
#if LIST_DIAGS
{
TICKS t2 = GetClockTicks();
unsigned uArea = RL.GetDPArea();
Log("ticks=%ld\n", (long) (t2 - t1));
Log("area=%u\n", uArea);
}
#endif
g_dDPAreaWithoutDiags += uLengthA*uLengthB;
double dDPAreaWithDiags = 0.0;
const unsigned uRegionCount = RL.GetCount();
for (unsigned uRegionIndex = 0; uRegionIndex < uRegionCount; ++uRegionIndex)
{
const DPRegion &r = RL.Get(uRegionIndex);
PWPath RegPath;
if (DPREGIONTYPE_Diag == r.m_Type)
{
DiagToPath(r.m_Diag, RegPath);
#if TRACE_PATH
Log("DiagToPath, path=\n");
RegPath.LogMe();
#endif
}
else if (DPREGIONTYPE_Rect == r.m_Type)
{
const unsigned uRegStartPosA = r.m_Rect.m_uStartPosA;
const unsigned uRegStartPosB = r.m_Rect.m_uStartPosB;
const unsigned uRegLengthA = r.m_Rect.m_uLengthA;
const unsigned uRegLengthB = r.m_Rect.m_uLengthB;
const ProfPos *RegPA = PA + uRegStartPosA;
const ProfPos *RegPB = PB + uRegStartPosB;
dDPAreaWithDiags += uRegLengthA*uRegLengthB;
GlobalAlignNoDiags(RegPA, uRegLengthA, RegPB, uRegLengthB, RegPath);
#if TRACE_PATH
Log("GlobalAlignNoDiags RegPath=\n");
RegPath.LogMe();
#endif
OffsetPath(RegPath, uRegStartPosA, uRegStartPosB);
#if TRACE_PATH
Log("After offset path, RegPath=\n");
RegPath.LogMe();
#endif
}
else
Quit("GlobalAlignDiags, Invalid region type %u", r.m_Type);
AppendRegPath(Path, RegPath);
#if TRACE_PATH
Log("After AppendPath, path=");
Path.LogMe();
#endif
}
#if TRACE
{
double dDPAreaWithoutDiags = uLengthA*uLengthB;
Log("DP area with diags %.3g without %.3g pct saved %.3g %%\n",
dDPAreaWithDiags, dDPAreaWithoutDiags, (1.0 - dDPAreaWithDiags/dDPAreaWithoutDiags)*100.0);
}
#endif
g_dDPAreaWithDiags += dDPAreaWithDiags;
return 0;
}
