obitools/obitools4
2
0
Fork 0
mirror of https://github.com/metabarcoding/obitools4.git synced 2026-08-24 13:51:18 +00:00
Code Issues Packages Projects Releases Wiki Activity

fix: validate inputs and fix coordinate offsets in pattern matching

Browse Source 
Introduce explicit input validation and a `(-1, -1, -1)` sentinel return value to signal invalid or unreliable matches. Add pre-call length checks and early returns to prevent panics during indel relocation. Fix an index offset bug in coordinate calculation by preserving the original fragment position before applying alignment deltas. Update match filtering to enforce reliability checks only on successfully relocated alignments. Comprehensive tests validate edge cases, boundary constraints, and error thresholds.
This commit is contained in:
Eric Coissac
2026-08-19 17:25:44 +02:00
parent 438893d910
commit 620646d412
3 changed files with 174 additions and 27 deletions
Download Patch File Download Diff File Expand all files Collapse all files
pkg/obiapat/pattern.go
+28 -25
View File
@@ -373,17 +373,7 @@ func (pattern ApatPattern) BestMatch(sequence ApatSequence, begin, length int) (
cpattern := (*[1 << 30]byte)(unsafe.Pointer(pattern.pointer.pointer.cpat))
frg := sequence.pointer.reference.Sequence()[start:end]
if len(frg) <= int(pattern.pointer.pointer.patlen) {
// The sequence is too short (e.g. the match is near one of its
// ends) to extract a fragment longer than the pattern, which
// obialign.LocatePattern requires. Keep the original match
// instead of refining it.
start = best[0]
end = best[1]
log.Debugln("Fragment too short for indel relocation, keeping original match", start, end, nerr)
return
}
fragStart := start
log.Debugln(
string(frg),
@@ -395,11 +385,20 @@ func (pattern ApatPattern) BestMatch(sequence ApatSequence, begin, length int) (
(*cpattern)[0:int(pattern.pointer.pointer.patlen)],
frg)
// olderr := m[2]
if from < 0 {
// obialign.LocatePattern could not reconstruct a reliable
// alignment (e.g. the fragment is too short relative to the
// pattern). Reporting a guessed position would risk placing
// the primer boundary incorrectly, so treat it as no match
// at all rather than falling back to an unrefined position.
matched = false
log.Debugln("No reliable indel relocation, discarding match", sequence.pointer.reference.Id())
return
}
nerr = score
start = start + from
end = start + to
start = fragStart + from
end = fragStart + to
log.Debugf("BestMatch on %s : score=%d [%d..%d]", sequence.pointer.reference.Id(), score, start, nerr)
return
}
@@ -478,6 +477,7 @@ func (pattern ApatPattern) AllMatches(sequence ApatSequence, begin, length int)
for _, m := range res {
// Recompute the start and end position of the match
// when the pattern allows for indels
valid := true
if m[2] > 0 && pattern.pointer.pointer.hasIndel {
// obilog.Warnf("Locating indel on sequence %s[%s]", sequence.pointer.reference.Id(), pattern.String())
start := m[0] - m[2]*2
@@ -491,17 +491,20 @@ func (pattern ApatPattern) AllMatches(sequence ApatSequence, begin, length int)
cpattern := (*[1 << 30]byte)(unsafe.Pointer(pattern.pointer.pointer.cpat))
frg := sequence.pointer.reference.Sequence()[start:end]
// obialign.LocatePattern requires the fragment to be strictly
// longer than the pattern. When the match sits near one of the
// sequence ends, the fragment can be clamped to the sequence
// boundaries and end up too short; in that case keep the
// original (unrefined) match instead of crashing.
if len(frg) > int(pattern.pointer.pointer.patlen) {
pb, pe, score := obialign.LocatePattern(
sequence.pointer.reference.Id(),
(*cpattern)[0:int(pattern.pointer.pointer.patlen)],
frg)
pb, pe, score := obialign.LocatePattern(
sequence.pointer.reference.Id(),
(*cpattern)[0:int(pattern.pointer.pointer.patlen)],
frg)
if pb < 0 {
// obialign.LocatePattern could not reconstruct a
// reliable alignment (e.g. the match sits too close
// to a sequence end for the fragment to be usable).
// Reporting a guessed position risks placing the
// primer boundary incorrectly, so drop the match
// entirely instead of keeping an unrefined guess.
valid = false
} else {
// olderr := m[2]
m[2] = score
m[0] = start + pb
@@ -512,7 +515,7 @@ func (pattern ApatPattern) AllMatches(sequence ApatSequence, begin, length int)
}
}
if int(pattern.pointer.pointer.maxerr) >= m[2] {
if valid && int(pattern.pointer.pointer.maxerr) >= m[2] {
res[j] = m
j++
}