|
| 1 | +package org.biojava.nbio.core.sequence.io; |
| 2 | + |
| 3 | +import org.biojava.nbio.core.sequence.ProteinSequence; |
| 4 | +import org.biojava.nbio.core.sequence.compound.AminoAcidCompound; |
| 5 | +import org.biojava.nbio.core.sequence.compound.AminoAcidCompoundSet; |
| 6 | +import org.biojava.nbio.core.sequence.io.template.SequenceCreatorInterface; |
| 7 | +import org.biojava.nbio.core.sequence.io.template.SequenceHeaderParserInterface; |
| 8 | +import org.biojava.nbio.core.util.InputStreamProvider; |
| 9 | + |
| 10 | +import java.io.File; |
| 11 | +import java.io.IOException; |
| 12 | +import java.io.InputStream; |
| 13 | +import java.io.UncheckedIOException; |
| 14 | +import java.nio.file.Path; |
| 15 | +import java.util.Collections; |
| 16 | +import java.util.Iterator; |
| 17 | +import java.util.LinkedHashMap; |
| 18 | +import java.util.Map; |
| 19 | +import java.util.Optional; |
| 20 | +import java.util.Spliterator; |
| 21 | +import java.util.Spliterators; |
| 22 | +import java.util.function.Consumer; |
| 23 | +import java.util.stream.Stream; |
| 24 | +import java.util.stream.StreamSupport; |
| 25 | + |
| 26 | +/** |
| 27 | + * Read from a FASTA file (or gzipped FASTA file) and create a Java stream of {@link ProteinSequence} objects |
| 28 | + * for use in a functional programming paradigm. |
| 29 | + * |
| 30 | + * @author Gary Murphy |
| 31 | + * @since 7.1.0 |
| 32 | + */ |
| 33 | +public class FastaStreamer { |
| 34 | + |
| 35 | + private final Path path; |
| 36 | + private int batchSize = 1_000; |
| 37 | + private SequenceHeaderParserInterface<ProteinSequence, AminoAcidCompound> headerParser; |
| 38 | + private SequenceCreatorInterface<AminoAcidCompound> sequenceCreator; |
| 39 | + private LinkedHashMap<String, ProteinSequence> chunk = new LinkedHashMap<>(); |
| 40 | + private Iterator<Map.Entry<String, ProteinSequence>> iterator = Collections.emptyIterator(); |
| 41 | + private boolean closed = false; |
| 42 | + |
| 43 | + /** |
| 44 | + * The constructor is private. Created via the <tt>from(...)</tt> static factory method |
| 45 | + * |
| 46 | + * @param path the path to the file containing the FASTA content (possibly GZipped) |
| 47 | + */ |
| 48 | + private FastaStreamer(final Path path) { |
| 49 | + this.path = path; |
| 50 | + } |
| 51 | + |
| 52 | + public static FastaStreamer from(final Path path) { |
| 53 | + return new FastaStreamer(path); |
| 54 | + } |
| 55 | + |
| 56 | + public static FastaStreamer from(File file) { |
| 57 | + return from(file.toPath()); |
| 58 | + } |
| 59 | + |
| 60 | + public FastaStreamer withHeaderParser(SequenceHeaderParserInterface<ProteinSequence, AminoAcidCompound> headerParser) { |
| 61 | + this.headerParser = headerParser; |
| 62 | + return this; |
| 63 | + } |
| 64 | + |
| 65 | + public FastaStreamer withSequenceCreator(SequenceCreatorInterface<AminoAcidCompound> sequenceCreator) { |
| 66 | + this.sequenceCreator = sequenceCreator; |
| 67 | + return this; |
| 68 | + } |
| 69 | + |
| 70 | + public FastaStreamer batchSize(int size) { |
| 71 | + this.batchSize = size; |
| 72 | + return this; |
| 73 | + } |
| 74 | + |
| 75 | + /** |
| 76 | + * Enable iteration through the proteins in the file using syntax such as: |
| 77 | + * <pre> |
| 78 | + * for(ProteinSequence sequence : FastaStreamer.from(path).each()) { |
| 79 | + * . |
| 80 | + * . |
| 81 | + * . |
| 82 | + * } |
| 83 | + * </pre> |
| 84 | + * |
| 85 | + * @return an iterable suitable for an iteration loop |
| 86 | + */ |
| 87 | + public Iterable<ProteinSequence> each() { |
| 88 | + return () -> stream().iterator(); |
| 89 | + } |
| 90 | + |
| 91 | + /** |
| 92 | + * Create a stream of protein sequences from the contents of the path |
| 93 | + * @return the stream |
| 94 | + */ |
| 95 | + public Stream<ProteinSequence> stream() { |
| 96 | + InputStreamProvider provider = new InputStreamProvider(); |
| 97 | + InputStream input; |
| 98 | + try { |
| 99 | + input = provider.getInputStream(getPath().toFile()); |
| 100 | + } catch (IOException exception) { |
| 101 | + throw new UncheckedIOException(exception); |
| 102 | + } |
| 103 | + FastaReader<ProteinSequence, AminoAcidCompound> reader = new FastaReader<>(input, getHeaderParser(), getSequenceCreator()); |
| 104 | + Spliterator<ProteinSequence> source = new Spliterators.AbstractSpliterator<>(Integer.MAX_VALUE, Spliterator.IMMUTABLE | Spliterator.NONNULL) { |
| 105 | + @Override |
| 106 | + public boolean tryAdvance(Consumer<? super ProteinSequence> action) { |
| 107 | + if (closed) { |
| 108 | + return false; |
| 109 | + } |
| 110 | + ProteinSequence protein = next(reader); |
| 111 | + if (null == protein) { |
| 112 | + return false; |
| 113 | + } |
| 114 | + action.accept(protein); |
| 115 | + return true; |
| 116 | + } |
| 117 | + |
| 118 | + /** |
| 119 | + * Fetch the next header/protein tuple from the cache. If the cache is empty, fetch another |
| 120 | + * batch from the source file |
| 121 | + * |
| 122 | + * @param reader |
| 123 | + * the input stream from which the FASTA content is read |
| 124 | + * @return the protein sequence |
| 125 | + */ |
| 126 | + private ProteinSequence next(FastaReader<ProteinSequence, AminoAcidCompound> reader) { |
| 127 | + try { |
| 128 | + if (!iterator.hasNext()) { |
| 129 | + chunk = reader.process(getBatchSize()); |
| 130 | + if (null == chunk) { |
| 131 | + closed = true; |
| 132 | + reader.close(); |
| 133 | + return null; |
| 134 | + } |
| 135 | + iterator = chunk.entrySet().iterator(); |
| 136 | + } |
| 137 | + if (iterator.hasNext()) { |
| 138 | + Map.Entry<String, ProteinSequence> entry = iterator.next(); |
| 139 | + return createSequence(entry.getValue()); |
| 140 | + } |
| 141 | + closed = true; |
| 142 | + reader.close(); |
| 143 | + } catch (IOException exception) { |
| 144 | + throw new UncheckedIOException(String.format("I/O error reading the FASTA file from '%s'", getPath()), exception); |
| 145 | + } |
| 146 | + return null; |
| 147 | + } |
| 148 | + }; // Spliterator |
| 149 | + return StreamSupport.stream(source, false); |
| 150 | + } |
| 151 | + |
| 152 | + /** |
| 153 | + * Create the sequence with the information from the header. This implementation return the sequence as-is, but |
| 154 | + * this is an opportunity for the implementer to build specific information into the user collection space |
| 155 | + * of the sequence |
| 156 | + * |
| 157 | + * @param sequence the protein sequence |
| 158 | + * @return the sequence |
| 159 | + */ |
| 160 | + protected ProteinSequence createSequence(ProteinSequence sequence) { |
| 161 | + return sequence; |
| 162 | + } |
| 163 | + |
| 164 | + protected Path getPath() { |
| 165 | + return path; |
| 166 | + } |
| 167 | + |
| 168 | + protected int getBatchSize() { |
| 169 | + return batchSize; |
| 170 | + } |
| 171 | + |
| 172 | + protected SequenceHeaderParserInterface<ProteinSequence, AminoAcidCompound> getHeaderParser() { |
| 173 | + return Optional.ofNullable(headerParser).orElse(new GenericFastaHeaderParser<>()); |
| 174 | + } |
| 175 | + |
| 176 | + public SequenceCreatorInterface<AminoAcidCompound> getSequenceCreator() { |
| 177 | + return Optional.ofNullable(sequenceCreator).orElse(new ProteinSequenceCreator(AminoAcidCompoundSet.getAminoAcidCompoundSet())); |
| 178 | + } |
| 179 | +} |
0 commit comments