void ComponentRepeat::notePositions(GlushkovBuildState &bs) { assert(m_max > 0); assert(m_max == NoLimit || m_max < MAX_MAX_BOUND); /* Note: We can construct smaller subgraphs if we're not maintaining edge * priorities. */ // We create one copy only through a recursive call to notePositions(), // first() and last(). Then we clone its positions and store the // appropriate firsts and lasts values for the copies. posFirst = bs.getBuilder().numVertices(); sub_comp->notePositions(bs); u32 copies = m_max < NoLimit ? m_max : MAX(m_min, 1); DEBUG_PRINTF("building %u copies of repeated region\n", copies); m_firsts.clear(); m_lasts.clear(); m_firsts.resize(copies); m_lasts.resize(copies); m_firsts[0] = sub_comp->first(); m_lasts[0] = sub_comp->last(); postSubNotePositionHook(); posLast = bs.getBuilder().numVertices() - 1; u32 vcount = posLast + 1 - posFirst; // If we're making more than one copy, then our firsts and lasts must only // contain vertices inside [posFirst, posLast]: anything else means we have // an embedded anchor or otherwise weird situation. if (copies > 1) { checkPositions(m_firsts[0], bs); checkPositions(m_lasts[0], bs); } // Avoid enormous expansions if (vcount * copies > MAX_POSITIONS_EXPANDED) { throw ParseError("Bounded repeat is too large."); } // Add positions for the rest of the copies size_t copyPositions = vcount * (copies - 1); bs.getBuilder().makePositions(copyPositions); // Calculate our firsts and lasts for the copies for (u32 i = 1; i < copies; ++i) { m_firsts[i] = m_firsts[0]; m_lasts[i] = m_lasts[0]; u32 base = i * vcount; addBase(base, m_firsts[i], m_lasts[i]); } recordPosBounds(posFirst, bs.getBuilder().numVertices()); precalc_firsts(); /* ComponentRepeat requires firsts to be calculated ahead * of time and cached due to expense */ }
void AsciiComponentClass::notePositions(GlushkovBuildState &bs) { // We should always be finalized by now. assert(finalized); NFABuilder &builder = bs.getBuilder(); position = builder.makePositions(1); builder.addCharReach(position, cr); builder.setNodeReportID(position, 0 /* offset adj */); recordPosBounds(position, position + 1); }