Merge pull request #71 from smithlabcode/self-overlapping-adapter-con…

…taminant-fix

Changing adapter matching so that self-overlapping patterns are only matched once per read

src/StreamReader.cpp

@@ -344,9 +344,10 @@ StreamReader::process_sequence_base_from_buffer(FastqStats &stats) {
       if (do_adapter_optimized && (num_bases_after_n == adapter_size)) {
         cur_kmer &= adapter_mask;
         for (size_t i = 0; i != num_adapters; ++i) {
-          if (cur_kmer == adapters[i]) {
+          if (cur_kmer == adapters[i] && !adapters_found[i]) {
             ++stats.pos_adapter_count[
               (read_pos << Constants::bit_shift_adapter) | i];
+            adapters_found[i] = true;
           }
         }
       }
@@ -436,6 +437,7 @@ StreamReader::read_sequence_line(FastqStats &stats) {
   still_in_buffer = true;
   next_truncation = 100;
   do_kmer_read = (stats.num_reads == next_kmer_read);
+  adapters_found.reset();
 
   if (do_adapters_slow) {
     const string seq_line_str = cur_char;

src/StreamReader.hpp

@@ -18,6 +18,7 @@
 
 #include <string>
 #include <cmath>
+#include <bitset>
 
 // Optional zlib usage
 #include <zlib.h>
@@ -81,6 +82,7 @@ class StreamReader{
   const size_t adapter_size;
   const size_t adapter_mask;
   const std::array<size_t, Constants::max_adapters> adapters;
+  std::bitset<Constants::max_adapters> adapters_found{};
 
   const std::string filename;
 
@@ -250,7 +252,7 @@ class BamReader : public StreamReader {
   bool read_entry(FastqStats &stats, size_t &num_bytes_read);
 
   // Specially made for BamReader to work directly with bam1_t
-  inline void read_sequence_line(FastqStats &stats);  
+  inline void read_sequence_line(FastqStats &stats);
   inline void read_quality_line(FastqStats &stats);  // parse quality
   inline void put_base_in_buffer(const size_t pos);  // puts base in buffer or leftover
   ~BamReader();

test/md5sum.txt

@@ -1,30 +1,30 @@
-9641ae08f984bde102d4292bcad56484  test_output/SRR1853178_1/fastqc_data.txt
+5c4e6118d438b1f01cf620120f26c622  test_output/SRR1853178_1/fastqc_data.txt
 36df1dcab539ba4ef885239fc8524636  test_output/SRR1853178_1/summary.txt
-c5b1d892705ee40353a5b88df5fc3a74  test_output/SRR3897196_2/fastqc_data.txt
+8dfecfc49e5b450152be12c63177054f  test_output/SRR3897196_2/fastqc_data.txt
 80cd130958bcb2219f1e5a07d06a3b6e  test_output/SRR3897196_2/summary.txt
-d60663e26511968865b3ed92a864fda8  test_output/SRR9624732_1/fastqc_data.txt
+5df2ae98f4389616df1fa90fe46ce463  test_output/SRR9624732_1/fastqc_data.txt
 a525d455b17eb6ba84cd2a60d281a2b0  test_output/SRR9624732_1/summary.txt
-053bb7a28541ac110116086ac6ea0cac  test_output/SRR10143153_2/fastqc_data.txt
+6e5eaadf209c3f77ab61ed61a034f49e  test_output/SRR10143153_2/fastqc_data.txt
 19f1811f324e4c44154f659bb6e22806  test_output/SRR10143153_2/summary.txt
-71e3964a9a610ebcb3207ea2a391b4be  test_output/SRR1772703.lite.1/fastqc_data.txt
-6403659cb0295ec05db9891e3e5e4f76  test_output/SRR1772703.lite.1/summary.txt
-8d28c215efc4f2930a002c240fb14448  test_output/SRR9624732_2/fastqc_data.txt
+427099afa91d877f078f1e7989fe4c39  test_output/SRR1772703.lite.1/fastqc_data.txt
+ad5727295e7c8de6eb6874837bf1518f  test_output/SRR1772703.lite.1/summary.txt
+04f7bb98120971c0ba1d648fda893a7c  test_output/SRR9624732_2/fastqc_data.txt
 fefc5d746f853c14b5e00421ad1ec260  test_output/SRR9624732_2/summary.txt
-3942d27ab09f2db7f58a2ed4c904ac8a  test_output/SRR10124060/fastqc_data.txt
-1e228b1bb498eec2ca81ec2cc657a02d  test_output/SRR10124060/summary.txt
-7c5f40be6a37ac4d99c39400b1f7ddb6  test_output/SRR891268_2/fastqc_data.txt
+a7827fe2115ef6ac48cd61f2065e58e5  test_output/SRR10124060/fastqc_data.txt
+776f7d1b53bbed8683de9ca1d2529f1e  test_output/SRR10124060/summary.txt
+7c05da833c8806ea8d5093ddab337f1c  test_output/SRR891268_2/fastqc_data.txt
 20a8e50baace4c672622793874a3d7de  test_output/SRR891268_2/summary.txt
-e02e6043667560aacb39b4e956edc146  test_output/SRR9878537.lite.1/fastqc_data.txt
+9bae57f4fa64d9fca4b11f0e0c107327  test_output/SRR9878537.lite.1/fastqc_data.txt
 e5c40997d4993c07e164ee5598c39cf9  test_output/SRR9878537.lite.1/summary.txt
-7807c0aeca3bf4fdb18309ce993f6e35  test_output/SRR891268_1/fastqc_data.txt
+6778518c16860c4300ed575d1dfdd43e  test_output/SRR891268_1/fastqc_data.txt
 69e7d0c53cd2e67117637c408b65333a  test_output/SRR891268_1/summary.txt
-54a383a9dd615f4130cbdab202829173  test_output/SRR6059706.lite.1/fastqc_data.txt
-a5545633e81fc57fd03ff6ff7ba4fb8f  test_output/SRR6059706.lite.1/summary.txt
-21b11f43971f424267f6e32a880aaea9  test_output/SRR6387347/fastqc_data.txt
+f05047b2a21949570ca2ae45b73b8da0  test_output/SRR6059706.lite.1/fastqc_data.txt
+e348e4bcc7fc6f05e989ac7858d2b287  test_output/SRR6059706.lite.1/summary.txt
+db82cb634e7e9b2f30472a069ba082b0  test_output/SRR6387347/fastqc_data.txt
 a61f65047e76f93300967cf399d044de  test_output/SRR6387347/summary.txt
-65519f6bcfd726b3219d7f646ad1d3b5  test_output/SRR3897196_1/fastqc_data.txt
+905dcc1d5d135e18df379cc1edf778b6  test_output/SRR3897196_1/fastqc_data.txt
 b736ee95d5c450ef5c0dda31957b6818  test_output/SRR3897196_1/summary.txt
-109cf880b376fcfba44b9637dcc045a6  test_output/SRR10143153_1/fastqc_data.txt
+f7eb7940ad9836764b695d2abb0426da  test_output/SRR10143153_1/fastqc_data.txt
 9ad191925d47a57d4f8b12f21ba0a7c3  test_output/SRR10143153_1/summary.txt
-dfebd8cc85b17f954fb14287a2dc8169  test_output/SRR1853178_2/fastqc_data.txt
-7645f17be12d149347dab6033aabdeb3  test_output/SRR1853178_2/summary.txt
+508cdda67a1ea7ddc756aa3656d1079b  test_output/SRR1853178_2/fastqc_data.txt
+c331d0f7a6aa9d72be41ac531f9ba269  test_output/SRR1853178_2/summary.txt