Scanner.py

# Copyright 2007-2017 by Peter Cock.  All rights reserved.
# Revisions copyright 2010 by Uri Laserson.  All rights reserved.
# This code is part of the Biopython distribution and governed by its
# license.  Please see the LICENSE file that should have been included
# as part of this package.
"""Internal code for parsing GenBank and EMBL files (PRIVATE).

This code is NOT intended for direct use.  It provides a basic scanner
(for use with a event consumer such as Bio.GenBank._FeatureConsumer)
to parse a GenBank or EMBL file (with their shared INSDC feature table).

It is used by Bio.GenBank to parse GenBank files
It is also used by Bio.SeqIO to parse GenBank and EMBL files

Feature Table Documentation:

- http://www.insdc.org/files/feature_table.html
- http://www.ncbi.nlm.nih.gov/projects/collab/FT/index.html
- ftp://ftp.ncbi.nih.gov/genbank/docs/
"""
# 17-MAR-2009: added wgs, wgs_scafld for GenBank whole genome shotgun master records.
# These are GenBank files that summarize the content of a project, and provide lists of
# scaffold and contig files in the project. These will be in annotations['wgs'] and
# annotations['wgs_scafld']. These GenBank files do not have sequences. See
# http://groups.google.com/group/bionet.molbio.genbank/browse_thread/thread/51fb88bf39e7dc36
# http://is.gd/nNgk
# for more details of this format, and an example.
# Added by Ying Huang & Iddo Friedberg


import warnings
import re
import sys
from collections import defaultdict

from Bio.File import as_handle
from Bio.Seq import Seq
from Bio.SeqRecord import SeqRecord
from Bio import BiopythonParserWarning


class InsdcScanner:
    """Basic functions for breaking up a GenBank/EMBL file into sub sections.

    The International Nucleotide Sequence Database Collaboration (INSDC)
    between the DDBJ, EMBL, and GenBank.  These organisations all use the
    same "Feature Table" layout in their plain text flat file formats.

    However, the header and sequence sections of an EMBL file are very
    different in layout to those produced by GenBank/DDBJ.
    """

    # These constants get redefined with sensible values in the sub classes:
    RECORD_START = "XXX"  # "LOCUS       " or "ID   "
    HEADER_WIDTH = 3  # 12 or 5
    FEATURE_START_MARKERS = ["XXX***FEATURES***XXX"]
    FEATURE_END_MARKERS = ["XXX***END FEATURES***XXX"]
    FEATURE_QUALIFIER_INDENT = 0
    FEATURE_QUALIFIER_SPACER = ""
    SEQUENCE_HEADERS = ["XXX"]  # with right hand side spaces removed

    def __init__(self, debug=0):
        """Initialize the class."""
        assert len(self.RECORD_START) == self.HEADER_WIDTH
        for marker in self.SEQUENCE_HEADERS:
            assert marker == marker.rstrip()
        assert len(self.FEATURE_QUALIFIER_SPACER) == self.FEATURE_QUALIFIER_INDENT
        self.debug = debug
        self.handle = None
        self.line = None

    def set_handle(self, handle):
        """Set the handle attribute."""
        self.handle = handle
        self.line = ""

    def find_start(self):
        """Read in lines until find the ID/LOCUS line, which is returned.

        Any preamble (such as the header used by the NCBI on ``*.seq.gz`` archives)
        will we ignored.
        """
        while True:
            if self.line:
                line = self.line
                self.line = ""
            else:
                line = self.handle.readline()
            if not line:
                if self.debug:
                    print("End of file")
                return None
            if isinstance(line[0], int):
                # Same exception as for FASTQ files
                raise ValueError("Is this handle in binary mode not text mode?")
            if line[: self.HEADER_WIDTH] == self.RECORD_START:
                if self.debug > 1:
                    print("Found the start of a record:\n" + line)
                break
            line = line.rstrip()
            if line == "//":
                if self.debug > 1:
                    print("Skipping // marking end of last record")
            elif line == "":
                if self.debug > 1:
                    print("Skipping blank line before record")
            else:
                # Ignore any header before the first ID/LOCUS line.
                if self.debug > 1:
                    print("Skipping header line before record:\n" + line)
        self.line = line
        return line

    def parse_header(self):
        """Return list of strings making up the header.

        New line characters are removed.

        Assumes you have just read in the ID/LOCUS line.
        """
        if self.line[: self.HEADER_WIDTH] != self.RECORD_START:
            raise ValueError("Not at start of record")

        header_lines = []
        while True:
            line = self.handle.readline()
            if not line:
                raise ValueError("Premature end of line during sequence data")
            line = line.rstrip()
            if line in self.FEATURE_START_MARKERS:
                if self.debug:
                    print("Found feature table")
                break
            # if line[:self.HEADER_WIDTH]==self.FEATURE_START_MARKER[:self.HEADER_WIDTH]:
            #    if self.debug : print("Found header table (?)")
            #    break
            if line[: self.HEADER_WIDTH].rstrip() in self.SEQUENCE_HEADERS:
                if self.debug:
                    print("Found start of sequence")
                break
            if line == "//":
                raise ValueError("Premature end of sequence data marker '//' found")
            header_lines.append(line)
        self.line = line
        return header_lines

    def parse_features(self, skip=False):
        """Return list of tuples for the features (if present).

        Each feature is returned as a tuple (key, location, qualifiers)
        where key and location are strings (e.g. "CDS" and
        "complement(join(490883..490885,1..879))") while qualifiers
        is a list of two string tuples (feature qualifier keys and values).

        Assumes you have already read to the start of the features table.
        """
        if self.line.rstrip() not in self.FEATURE_START_MARKERS:
            if self.debug:
                print("Didn't find any feature table")
            return []

        while self.line.rstrip() in self.FEATURE_START_MARKERS:
            self.line = self.handle.readline()

        features = []
        line = self.line
        while True:
            if not line:
                raise ValueError("Premature end of line during features table")
            if line[: self.HEADER_WIDTH].rstrip() in self.SEQUENCE_HEADERS:
                if self.debug:
                    print("Found start of sequence")
                break
            line = line.rstrip()
            if line == "//":
                raise ValueError("Premature end of features table, marker '//' found")
            if line in self.FEATURE_END_MARKERS:
                if self.debug:
                    print("Found end of features")
                line = self.handle.readline()
                break
            if line[2 : self.FEATURE_QUALIFIER_INDENT].strip() == "":
                # This is an empty feature line between qualifiers. Empty
                # feature lines within qualifiers are handled below (ignored).
                line = self.handle.readline()
                continue
            if len(line) < self.FEATURE_QUALIFIER_INDENT:
                warnings.warn(
                    "line too short to contain a feature: %r" % line,
                    BiopythonParserWarning,
                )
                line = self.handle.readline()
                continue

            if skip:
                line = self.handle.readline()
                while (
                    line[: self.FEATURE_QUALIFIER_INDENT]
                    == self.FEATURE_QUALIFIER_SPACER
                ):
                    line = self.handle.readline()
            else:
                # Build up a list of the lines making up this feature:
                if (
                    line[self.FEATURE_QUALIFIER_INDENT] != " "
                    and " " in line[self.FEATURE_QUALIFIER_INDENT :]
                ):
                    # The feature table design enforces a length limit on the feature keys.
                    # Some third party files (e.g. IGMT's EMBL like files) solve this by
                    # over indenting the location and qualifiers.
                    feature_key, line = line[2:].strip().split(None, 1)
                    feature_lines = [line]
                    warnings.warn(
                        "Over indented %s feature?" % feature_key,
                        BiopythonParserWarning,
                    )
                else:
                    feature_key = line[2 : self.FEATURE_QUALIFIER_INDENT].strip()
                    feature_lines = [line[self.FEATURE_QUALIFIER_INDENT :]]
                line = self.handle.readline()
                while line[
                    : self.FEATURE_QUALIFIER_INDENT
                ] == self.FEATURE_QUALIFIER_SPACER or (
                    line != "" and line.rstrip() == ""
                ):  # cope with blank lines in the midst of a feature
                    # Use strip to remove any harmless trailing white space AND and leading
                    # white space (e.g. out of spec files with too much indentation)
                    feature_lines.append(line[self.FEATURE_QUALIFIER_INDENT :].strip())
                    line = self.handle.readline()
                features.append(self.parse_feature(feature_key, feature_lines))
        self.line = line
        return features

    def parse_feature(self, feature_key, lines):
        r"""Parse a feature given as a list of strings into a tuple.

        Expects a feature as a list of strings, returns a tuple (key, location,
        qualifiers)

        For example given this GenBank feature::

             CDS             complement(join(490883..490885,1..879))
                             /locus_tag="NEQ001"
                             /note="conserved hypothetical [Methanococcus jannaschii];
                             COG1583:Uncharacterized ACR; IPR001472:Bipartite nuclear
                             localization signal; IPR002743: Protein of unknown
                             function DUF57"
                             /codon_start=1
                             /transl_table=11
                             /product="hypothetical protein"
                             /protein_id="NP_963295.1"
                             /db_xref="GI:41614797"
                             /db_xref="GeneID:2732620"
                             /translation="MRLLLELKALNSIDKKQLSNYLIQGFIYNILKNTEYSWLHNWKK
                             EKYFNFTLIPKKDIIENKRYYLIISSPDKRFIEVLHNKIKDLDIITIGLAQFQLRKTK
                             KFDPKLRFPWVTITPIVLREGKIVILKGDKYYKVFVKRLEELKKYNLIKKKEPILEEP
                             IEISLNQIKDGWKIIDVKDRYYDFRNKSFSAFSNWLRDLKEQSLRKYNNFCGKNFYFE
                             EAIFEGFTFYKTVSIRIRINRGEAVYIGTLWKELNVYRKLDKEEREFYKFLYDCGLGS
                             LNSMGFGFVNTKKNSAR"

        Then should give input key="CDS" and the rest of the data as a list of strings
        lines=["complement(join(490883..490885,1..879))", ..., "LNSMGFGFVNTKKNSAR"]
        where the leading spaces and trailing newlines have been removed.

        Returns tuple containing: (key as string, location string, qualifiers as list)
        as follows for this example:

        key = "CDS", string
        location = "complement(join(490883..490885,1..879))", string
        qualifiers = list of string tuples:

        [('locus_tag', '"NEQ001"'),
         ('note', '"conserved hypothetical [Methanococcus jannaschii];\nCOG1583:..."'),
         ('codon_start', '1'),
         ('transl_table', '11'),
         ('product', '"hypothetical protein"'),
         ('protein_id', '"NP_963295.1"'),
         ('db_xref', '"GI:41614797"'),
         ('db_xref', '"GeneID:2732620"'),
         ('translation', '"MRLLLELKALNSIDKKQLSNYLIQGFIYNILKNTEYSWLHNWKK\nEKYFNFT..."')]

        In the above example, the "note" and "translation" were edited for compactness,
        and they would contain multiple new line characters (displayed above as \n)

        If a qualifier is quoted (in this case, everything except codon_start and
        transl_table) then the quotes are NOT removed.

        Note that no whitespace is removed.
        """
        # Skip any blank lines
        iterator = (x for x in lines if x)
        try:
            line = next(iterator)

            feature_location = line.strip()
            while feature_location[-1:] == ",":
                # Multiline location, still more to come!
                line = next(iterator)
                feature_location += line.strip()
            if feature_location.count("(") > feature_location.count(")"):
                # Including the prev line in warning would be more explicit,
                # but this way get one-and-only-one warning shown by default:
                warnings.warn(
                    "Non-standard feature line wrapping (didn't break on comma)?",
                    BiopythonParserWarning,
                )
                while feature_location[-1:] == "," or feature_location.count(
                    "("
                ) > feature_location.count(")"):
                    line = next(iterator)
                    feature_location += line.strip()

            qualifiers = []

            for line_number, line in enumerate(iterator):
                # check for extra wrapping of the location closing parentheses
                if line_number == 0 and line.startswith(")"):
                    feature_location += line.strip()
                elif line[0] == "/":
                    # New qualifier
                    i = line.find("=")
                    key = line[1:i]  # does not work if i==-1
                    value = line[i + 1 :]  # we ignore 'value' if i==-1
                    if i and value.startswith(" ") and value.lstrip().startswith('"'):
                        warnings.warn(
                            "White space after equals in qualifier",
                            BiopythonParserWarning,
                        )
                        value = value.lstrip()
                    if i == -1:
                        # Qualifier with no key, e.g. /pseudo
                        key = line[1:]
                        qualifiers.append((key, None))
                    elif not value:
                        # ApE can output /note=
                        qualifiers.append((key, ""))
                    elif value == '"':
                        # One single quote
                        if self.debug:
                            print("Single quote %s:%s" % (key, value))
                        # DO NOT remove the quote...
                        qualifiers.append((key, value))
                    elif value[0] == '"':
                        # Quoted...
                        value_list = [value]
                        while value_list[-1][-1] != '"':
                            value_list.append(next(iterator))
                        value = "\n".join(value_list)
                        # DO NOT remove the quotes...
                        qualifiers.append((key, value))
                    else:
                        # Unquoted
                        # if debug : print("Unquoted line %s:%s" % (key,value))
                        qualifiers.append((key, value))
                else:
                    # Unquoted continuation
                    assert len(qualifiers) > 0
                    assert key == qualifiers[-1][0]
                    # if debug : print("Unquoted Cont %s:%s" % (key, line))
                    if qualifiers[-1][1] is None:
                        raise StopIteration
                    qualifiers[-1] = (key, qualifiers[-1][1] + "\n" + line)
            return feature_key, feature_location, qualifiers
        except StopIteration:
            # Bummer
            raise ValueError(
                "Problem with '%s' feature:\n%s" % (feature_key, "\n".join(lines))
            ) from None

    def parse_footer(self):
        """Return a tuple containing a list of any misc strings, and the sequence."""
        # This is a basic bit of code to scan and discard the sequence,
        # which was useful when developing the sub classes.
        if self.line in self.FEATURE_END_MARKERS:
            while self.line[: self.HEADER_WIDTH].rstrip() not in self.SEQUENCE_HEADERS:
                self.line = self.handle.readline()
                if not self.line:
                    raise ValueError("Premature end of file")
                self.line = self.line.rstrip()

        if self.line[: self.HEADER_WIDTH].rstrip() not in self.SEQUENCE_HEADERS:
            raise ValueError("Not at start of sequence")
        while True:
            line = self.handle.readline()
            if not line:
                raise ValueError("Premature end of line during sequence data")
            line = line.rstrip()
            if line == "//":
                break
        self.line = line
        return [], ""  # Dummy values!

    def _feed_first_line(self, consumer, line):
        """Handle the LOCUS/ID line, passing data to the comsumer (PRIVATE).

        This should be implemented by the EMBL / GenBank specific subclass

        Used by the parse_records() and parse() methods.
        """
        pass

    def _feed_header_lines(self, consumer, lines):
        """Handle the header lines (list of strings), passing data to the comsumer (PRIVATE).

        This should be implemented by the EMBL / GenBank specific subclass

        Used by the parse_records() and parse() methods.
        """
        pass

    @staticmethod
    def _feed_feature_table(consumer, feature_tuples):
        """Handle the feature table (list of tuples), passing data to the comsumer (PRIVATE).

        Used by the parse_records() and parse() methods.
        """
        consumer.start_feature_table()
        for feature_key, location_string, qualifiers in feature_tuples:
            consumer.feature_key(feature_key)
            consumer.location(location_string)
            for q_key, q_value in qualifiers:
                if q_value is None:
                    consumer.feature_qualifier(q_key, q_value)
                else:
                    consumer.feature_qualifier(q_key, q_value.replace("\n", " "))

    def _feed_misc_lines(self, consumer, lines):
        """Handle any lines between features and sequence (list of strings), passing data to the consumer (PRIVATE).

        This should be implemented by the EMBL / GenBank specific subclass

        Used by the parse_records() and parse() methods.
        """
        pass

    def feed(self, handle, consumer, do_features=True):
        """Feed a set of data into the consumer.

        This method is intended for use with the "old" code in Bio.GenBank

        Arguments:
         - handle - A handle with the information to parse.
         - consumer - The consumer that should be informed of events.
         - do_features - Boolean, should the features be parsed?
           Skipping the features can be much faster.

        Return values:
         - true  - Passed a record
         - false - Did not find a record

        """
        # Should work with both EMBL and GenBank files provided the
        # equivalent Bio.GenBank._FeatureConsumer methods are called...
        self.set_handle(handle)
        if not self.find_start():
            # Could not find (another) record
            consumer.data = None
            return False

        # We use the above class methods to parse the file into a simplified format.
        # The first line, header lines and any misc lines after the features will be
        # dealt with by GenBank / EMBL specific derived classes.

        # First line and header:
        self._feed_first_line(consumer, self.line)
        self._feed_header_lines(consumer, self.parse_header())

        # Features (common to both EMBL and GenBank):
        if do_features:
            self._feed_feature_table(consumer, self.parse_features(skip=False))
        else:
            self.parse_features(skip=True)  # ignore the data

        # Footer and sequence
        misc_lines, sequence_string = self.parse_footer()
        self._feed_misc_lines(consumer, misc_lines)

        consumer.sequence(sequence_string)
        # Calls to consumer.base_number() do nothing anyway
        consumer.record_end("//")

        assert self.line == "//"

        # And we are done
        return True

    def parse(self, handle, do_features=True):
        """Return a SeqRecord (with SeqFeatures if do_features=True).

        See also the method parse_records() for use on multi-record files.
        """
        from Bio.GenBank import _FeatureConsumer
        from Bio.GenBank.utils import FeatureValueCleaner

        consumer = _FeatureConsumer(
            use_fuzziness=1, feature_cleaner=FeatureValueCleaner()
        )

        if self.feed(handle, consumer, do_features):
            return consumer.data
        else:
            return None

    def parse_records(self, handle, do_features=True):
        """Parse records, return a SeqRecord object iterator.

        Each record (from the ID/LOCUS line to the // line) becomes a SeqRecord

        The SeqRecord objects include SeqFeatures if do_features=True

        This method is intended for use in Bio.SeqIO
        """
        # This is a generator function
        with as_handle(handle) as handle:
            while True:
                record = self.parse(handle, do_features)
                if record is None:
                    break
                if record.id is None:
                    raise ValueError(
                        "Failed to parse the record's ID. Invalid ID line?"
                    )
                if record.name == "<unknown name>":
                    raise ValueError(
                        "Failed to parse the record's name. Invalid ID line?"
                    )
                if record.description == "<unknown description>":
                    raise ValueError("Failed to parse the record's description")
                yield record

    def parse_cds_features(
        self, handle, alphabet=None, tags2id=("protein_id", "locus_tag", "product")
    ):
        """Parse CDS features, return SeqRecord object iterator.

        Each CDS feature becomes a SeqRecord.

        Arguments:
         - alphabet - Obsolete, should be left as None.
         - tags2id  - Tupple of three strings, the feature keys to use
           for the record id, name and description,

        This method is intended for use in Bio.SeqIO

        """
        if alphabet is not None:
            raise ValueError("The alphabet argument is no longer supported")
        with as_handle(handle) as handle:
            self.set_handle(handle)
            while self.find_start():
                # Got an EMBL or GenBank record...
                self.parse_header()  # ignore header lines!
                feature_tuples = self.parse_features()
                # self.parse_footer() # ignore footer lines!
                while True:
                    line = self.handle.readline()
                    if not line:
                        break
                    if line[:2] == "//":
                        break
                self.line = line.rstrip()

                # Now go though those features...
                for key, location_string, qualifiers in feature_tuples:
                    if key == "CDS":
                        # Create SeqRecord
                        # ================
                        # SeqRecord objects cannot be created with annotations, they
                        # must be added afterwards.  So create an empty record and
                        # then populate it:
                        record = SeqRecord(seq=None)
                        annotations = record.annotations
                        annotations["molecule_type"] = "protein"
                        # Should we add a location object to the annotations?
                        # I *think* that only makes sense for SeqFeatures with their
                        # sub features...
                        annotations["raw_location"] = location_string.replace(" ", "")

                        for (qualifier_name, qualifier_data) in qualifiers:
                            if (
                                qualifier_data is not None
                                and qualifier_data[0] == '"'
                                and qualifier_data[-1] == '"'
                            ):
                                # Remove quotes
                                qualifier_data = qualifier_data[1:-1]
                            # Append the data to the annotation qualifier...
                            if qualifier_name == "translation":
                                assert record.seq is None, "Multiple translations!"
                                record.seq = Seq(qualifier_data.replace("\n", ""))
                            elif qualifier_name == "db_xref":
                                # its a list, possibly empty.  Its safe to extend
                                record.dbxrefs.append(qualifier_data)
                            else:
                                if qualifier_data is not None:
                                    qualifier_data = qualifier_data.replace(
                                        "\n", " "
                                    ).replace("  ", " ")
                                try:
                                    annotations[qualifier_name] += " " + qualifier_data
                                except KeyError:
                                    # Not an addition to existing data, its the first bit
                                    annotations[qualifier_name] = qualifier_data

                        # Fill in the ID, Name, Description
                        # =================================
                        try:
                            record.id = annotations[tags2id[0]]
                        except KeyError:
                            pass
                        try:
                            record.name = annotations[tags2id[1]]
                        except KeyError:
                            pass
                        try:
                            record.description = annotations[tags2id[2]]
                        except KeyError:
                            pass

                        yield record


class EmblScanner(InsdcScanner):
    """For extracting chunks of information in EMBL files."""

    RECORD_START = "ID   "
    HEADER_WIDTH = 5
    FEATURE_START_MARKERS = ["FH   Key             Location/Qualifiers", "FH"]
    FEATURE_END_MARKERS = ["XX"]  # XX can also mark the end of many things!
    FEATURE_QUALIFIER_INDENT = 21
    FEATURE_QUALIFIER_SPACER = "FT" + " " * (FEATURE_QUALIFIER_INDENT - 2)
    SEQUENCE_HEADERS = ["SQ", "CO"]  # Remove trailing spaces

    EMBL_INDENT = HEADER_WIDTH
    EMBL_SPACER = " " * EMBL_INDENT

    def parse_footer(self):
        """Return a tuple containing a list of any misc strings, and the sequence."""
        if self.line[: self.HEADER_WIDTH].rstrip() not in self.SEQUENCE_HEADERS:
            raise ValueError("Footer format unexpected: '%s'" % self.line)

        # Note that the SQ line can be split into several lines...
        misc_lines = []
        while self.line[: self.HEADER_WIDTH].rstrip() in self.SEQUENCE_HEADERS:
            misc_lines.append(self.line)
            self.line = self.handle.readline()
            if not self.line:
                raise ValueError("Premature end of file")
            self.line = self.line.rstrip()

        if not (
            self.line[: self.HEADER_WIDTH] == " " * self.HEADER_WIDTH
            or self.line.strip() == "//"
        ):
            raise ValueError("Unexpected content after SQ or CO line: %r" % self.line)

        seq_lines = []
        line = self.line
        while True:
            if not line:
                raise ValueError("Premature end of file in sequence data")
            line = line.strip()
            if not line:
                raise ValueError("Blank line in sequence data")
            if line == "//":
                break
            if self.line[: self.HEADER_WIDTH] != (" " * self.HEADER_WIDTH):
                raise ValueError(
                    "Problem with characters in header line, "
                    " or incorrect header width: " + self.line
                )
            # Remove tailing number now, remove spaces later
            linersplit = line.rsplit(None, 1)
            if len(linersplit) == 2 and linersplit[1].isdigit():
                seq_lines.append(linersplit[0])
            elif line.isdigit():
                # Special case of final blank line with no bases
                # just the sequence coordinate
                pass
            else:
                warnings.warn(
                    "EMBL sequence line missing coordinates", BiopythonParserWarning
                )
                seq_lines.append(line)
            line = self.handle.readline()
        self.line = line
        return misc_lines, "".join(seq_lines).replace(" ", "")

    def _feed_first_line(self, consumer, line):
        assert line[: self.HEADER_WIDTH].rstrip() == "ID"
        if line[self.HEADER_WIDTH :].count(";") == 6:
            # Looks like the semi colon separated style introduced in 2006
            self._feed_first_line_new(consumer, line)
        elif line[self.HEADER_WIDTH :].count(";") == 3:
            if line.rstrip().endswith(" SQ"):
                # EMBL-bank patent data
                self._feed_first_line_patents(consumer, line)
            else:
                # Looks like the pre 2006 style
                self._feed_first_line_old(consumer, line)
        elif line[self.HEADER_WIDTH :].count(";") == 2:
            # Looks like KIKO patent data
            self._feed_first_line_patents_kipo(consumer, line)
        else:
            raise ValueError("Did not recognise the ID line layout:\n" + line)

    def _feed_first_line_patents(self, consumer, line):
        # Old style EMBL patent records where ID line ended SQ
        # Not 100% sure that PRT here is really molecule type and
        # not the data file division...
        #
        # Either Non-Redundant Level 1 database records,
        # ID <accession>; <molecule type>; <non-redundant level 1>; <cluster size L1>
        # e.g. ID   NRP_AX000635; PRT; NR1; 15 SQ
        #
        # Or, Non-Redundant Level 2 database records:
        # ID <L2-accession>; <molecule type>; <non-redundant level 2>; <cluster size L2>
        # e.g. ID   NRP0000016E; PRT; NR2; 5 SQ
        # e.g. ID   NRP_AX000635; PRT; NR1; 15 SQ
        fields = [
            data.strip() for data in line[self.HEADER_WIDTH :].strip()[:-3].split(";")
        ]
        assert len(fields) == 4
        consumer.locus(fields[0])
        consumer.residue_type(fields[1])  # semi-redundant
        consumer.data_file_division(fields[2])
        # TODO - Record cluster size?

    def _feed_first_line_patents_kipo(self, consumer, line):
        # EMBL format patent sequence from KIPO, e.g.
        # ftp://ftp.ebi.ac.uk/pub/databases/patentdata/kipo_prt.dat.gz
        #
        # e.g. ID   DI500001       STANDARD;      PRT;   111 AA.
        #
        # This follows the style of _feed_first_line_old
        assert line[: self.HEADER_WIDTH].rstrip() == "ID"
        fields = [line[self.HEADER_WIDTH :].split(None, 1)[0]]
        fields.extend(line[self.HEADER_WIDTH :].split(None, 1)[1].split(";"))
        fields = [entry.strip() for entry in fields]
        """
        The tokens represent:

           0. Primary accession number
           (space sep)
           1. ??? (e.g. standard)
           (semi-colon)
           2. Molecule type (protein)? Division? Always 'PRT'
           3. Sequence length (e.g. '111 AA.')
        """
        consumer.locus(fields[0])  # Should we also call the accession consumer?
        # consumer.molecule_type(fields[2])
        self._feed_seq_length(consumer, fields[3])

    def _feed_first_line_old(self, consumer, line):
        # Expects an ID line in the style before 2006, e.g.
        # ID   SC10H5 standard; DNA; PRO; 4870 BP.
        # ID   BSUB9999   standard; circular DNA; PRO; 4214630 BP.
        assert line[: self.HEADER_WIDTH].rstrip() == "ID"
        fields = [line[self.HEADER_WIDTH :].split(None, 1)[0]]
        fields.extend(line[self.HEADER_WIDTH :].split(None, 1)[1].split(";"))
        fields = [entry.strip() for entry in fields]
        """
        The tokens represent:

           0. Primary accession number
           (space sep)
           1. ??? (e.g. standard)
           (semi-colon)
           2. Topology and/or Molecule type (e.g. 'circular DNA' or 'DNA')
           3. Taxonomic division (e.g. 'PRO')
           4. Sequence length (e.g. '4639675 BP.')

        """
        consumer.locus(fields[0])  # Should we also call the accession consumer?
        consumer.residue_type(fields[2])
        if "circular" in fields[2]:
            consumer.topology("circular")
            consumer.molecule_type(fields[2].replace("circular", "").strip())
        elif "linear" in fields[2]:
            consumer.topology("linear")
            consumer.molecule_type(fields[2].replace("linear", "").strip())
        else:
            consumer.molecule_type(fields[2].strip())
        consumer.data_file_division(fields[3])
        self._feed_seq_length(consumer, fields[4])

    def _feed_first_line_new(self, consumer, line):
        # Expects an ID line in the style introduced in 2006, e.g.
        # ID   X56734; SV 1; linear; mRNA; STD; PLN; 1859 BP.
        # ID   CD789012; SV 4; linear; genomic DNA; HTG; MAM; 500 BP.
        assert line[: self.HEADER_WIDTH].rstrip() == "ID"
        fields = [data.strip() for data in line[self.HEADER_WIDTH :].strip().split(";")]
        assert len(fields) == 7
        """
        The tokens represent:

           0. Primary accession number
           1. Sequence version number
           2. Topology: 'circular' or 'linear'
           3. Molecule type (e.g. 'genomic DNA')
           4. Data class (e.g. 'STD')
           5. Taxonomic division (e.g. 'PRO')
           6. Sequence length (e.g. '4639675 BP.')

        """

        consumer.locus(fields[0])

        # Call the accession consumer now, to make sure we record
        # something as the record.id, in case there is no AC line
        consumer.accession(fields[0])

        # TODO - How to deal with the version field?  At the moment the consumer
        # will try and use this for the ID which isn't ideal for EMBL files.
        version_parts = fields[1].split()
        if (
            len(version_parts) == 2
            and version_parts[0] == "SV"
            and version_parts[1].isdigit()
        ):
            consumer.version_suffix(version_parts[1])

        # Based on how the old GenBank parser worked, merge these two:
        consumer.residue_type(" ".join(fields[2:4]))  # Semi-obsolete

        consumer.topology(fields[2])
        consumer.molecule_type(fields[3])

        # consumer.xxx(fields[4]) # TODO - What should we do with the data class?

        consumer.data_file_division(fields[5])

        self._feed_seq_length(consumer, fields[6])

    @staticmethod
    def _feed_seq_length(consumer, text):
        length_parts = text.split()
        assert len(length_parts) == 2, "Invalid sequence length string %r" % text
        assert length_parts[1].upper() in ["BP", "BP.", "AA", "AA."]
        consumer.size(length_parts[0])

    def _feed_header_lines(self, consumer, lines):
        consumer_dict = {
            "AC": "accession",
            "SV": "version",  # SV line removed in June 2006, now part of ID line
            "DE": "definition",
            # 'RN' : 'reference_num',
            # 'RC' : reference comment... TODO
            # 'RP' : 'reference_bases',
            # 'RX' : reference cross reference... DOI or Pubmed
            "RG": "consrtm",  # optional consortium
            # 'RA' : 'authors',
            # 'RT' : 'title',
            "RL": "journal",
            "OS": "organism",
            "OC": "taxonomy",
            # 'DR' : data reference
            "CC": "comment",
            # 'XX' : splitter
        }
        # We have to handle the following specially:
        # RX (depending on reference type...)
        for line in lines:
            line_type = line[: self.EMBL_INDENT].strip()
            data = line[self.EMBL_INDENT :].strip()
            if line_type == "XX":
                pass
            elif line_type == "RN":
                # Reformat reference numbers for the GenBank based consumer
                # e.g. '[1]' becomes '1'
                if data[0] == "[" and data[-1] == "]":
                    data = data[1:-1]
                consumer.reference_num(data)
            elif line_type == "RP":
                if data.strip() == "[-]":
                    # Patent EMBL files from KIPO just use: RN  [-]
                    pass
                else:
                    # Reformat reference numbers for the GenBank based consumer
                    # e.g. '1-4639675' becomes '(bases 1 to 4639675)'
                    # and '160-550, 904-1055' becomes '(bases 160 to 550; 904 to 1055)'
                    # Note could be multi-line, and end with a comma
                    parts = [
                        bases.replace("-", " to ").strip()
                        for bases in data.split(",")
                        if bases.strip()
                    ]
                    consumer.reference_bases("(bases %s)" % "; ".join(parts))
            elif line_type == "RT":
                # Remove the enclosing quotes and trailing semi colon.
                # Note the title can be split over multiple lines.
                if data.startswith('"'):
                    data = data[1:]
                if data.endswith('";'):
                    data = data[:-2]
                consumer.title(data)
            elif line_type == "RX":
                # EMBL support three reference types at the moment:
                # - PUBMED    PUBMED bibliographic database (NLM)
                # - DOI       Digital Object Identifier (International DOI Foundation)
                # - AGRICOLA  US National Agriculture Library (NAL) of the US Department
                #             of Agriculture (USDA)
                #
                # Format:
                # RX  resource_identifier; identifier.
                #
                # e.g.
                # RX   DOI; 10.1016/0024-3205(83)90010-3.
                # RX   PUBMED; 264242.
                #
                # Currently our reference object only supports PUBMED and MEDLINE
                # (as these were in GenBank files?).
                key, value = data.split(";", 1)
                if value.endswith("."):
                    value = value[:-1]
                value = value.strip()
                if key == "PUBMED":
                    consumer.pubmed_id(value)
                # TODO - Handle other reference types (here and in BioSQL bindings)
            elif line_type == "CC":
                # Have to pass a list of strings for this one (not just a string)
                consumer.comment([data])
            elif line_type == "DR":
                # Database Cross-reference, format:
                # DR   database_identifier; primary_identifier; secondary_identifier.
                #
                # e.g.
                # DR   MGI; 98599; Tcrb-V4.
                #
                # TODO - How should we store any secondary identifier?
                parts = data.rstrip(".").split(";")
                # Turn it into "database_identifier:primary_identifier" to
                # mimic the GenBank parser. e.g. "MGI:98599"
                if len(parts) == 1:
                    warnings.warn(
                        "Malformed DR line in EMBL file.", BiopythonParserWarning
                    )
                else:
                    consumer.dblink("%s:%s" % (parts[0].strip(), parts[1].strip()))
            elif line_type == "RA":
                # Remove trailing ; at end of authors list
                consumer.authors(data.rstrip(";"))
            elif line_type == "PR":
                # In the EMBL patent files, this is a PR (PRiority) line which
                # provides the earliest active priority within the family.
                # The priority  number comes first, followed by the priority date.
                #
                # e.g.
                # PR   JP19990377484 16-DEC-1999
                #
                # However, in most EMBL files this is a PR (PRoject) line which
                # gives the BioProject reference number.
                #
                # e.g.
                # PR   Project:PRJNA60715;
                #
                # In GenBank files this corresponds to the old PROJECT line
                # which was later replaced with the DBLINK line.
                if data.startswith("Project:"):
                    # Remove trailing ; at end of the project reference
                    consumer.project(data.rstrip(";"))
            elif line_type == "KW":
                consumer.keywords(data.rstrip(";"))
            elif line_type in consumer_dict:
                # Its a semi-automatic entry!
                getattr(consumer, consumer_dict[line_type])(data)
            else:
                if self.debug:
                    print("Ignoring EMBL header line:\n%s" % line)

    def _feed_misc_lines(self, consumer, lines):
        # TODO - Should we do something with the information on the SQ line(s)?
        lines.append("")
        line_iter = iter(lines)
        try:
            for line in line_iter:
                if line.startswith("CO   "):
                    line = line[5:].strip()
                    contig_location = line
                    while True:
                        line = next(line_iter)
                        if not line:
                            break
                        elif line.startswith("CO   "):
                            # Don't need to preseve the whitespace here.
                            contig_location += line[5:].strip()
                        else:
                            raise ValueError(
                                "Expected CO (contig) continuation line, got:\n" + line
                            )
                    consumer.contig_location(contig_location)
                if line.startswith("SQ   Sequence "):
                    # e.g.
                    # SQ   Sequence 219 BP; 82 A; 48 C; 33 G; 45 T; 11 other;
                    #
                    # Or, EMBL-bank patent, e.g.
                    # SQ   Sequence 465 AA; 3963407aa91d3a0d622fec679a4524e0; MD5;
                    self._feed_seq_length(
                        consumer, line[14:].rstrip().rstrip(";").split(";", 1)[0]
                    )
                    # TODO - Record the checksum etc?
            return
        except StopIteration:
            raise ValueError("Problem in misc lines before sequence") from None


class _ImgtScanner(EmblScanner):
    """For extracting chunks of information in IMGT (EMBL like) files (PRIVATE).

    IMGT files are like EMBL files but in order to allow longer feature types
    the features should be indented by 25 characters not 21 characters. In
    practice the IMGT flat files tend to use either 21 or 25 characters, so we
    must cope with both.

    This is private to encourage use of Bio.SeqIO rather than Bio.GenBank.
    """

    FEATURE_START_MARKERS = [
        "FH   Key             Location/Qualifiers",
        "FH   Key             Location/Qualifiers (from EMBL)",
        "FH   Key                 Location/Qualifiers",
        "FH",
    ]

    def _feed_first_line(self, consumer, line):
        assert line[: self.HEADER_WIDTH].rstrip() == "ID"
        if line[self.HEADER_WIDTH :].count(";") != 5:
            # Assume its an older EMBL-like line,
            return EmblScanner._feed_first_line(self, consumer, line)
        # Otherwise assume its the new (circa 2016) IMGT style
        # as used in the IPD-IMGT/HLA Database
        #
        # https://github.com/ANHIG/IMGTHLA/
        #
        # The key changes post 3.16 are the addition of an SV value
        # to the ID line, these additions should make the format more
        # similar to the ENA style.
        #
        # ID   HLA00001   standard; DNA; HUM; 3503 BP.
        #
        # becomes
        #
        # ID   HLA00001; SV 1; standard; DNA; HUM; 3503 BP.
        fields = [data.strip() for data in line[self.HEADER_WIDTH :].strip().split(";")]
        assert len(fields) == 6
        """
        The tokens represent:

           0. Primary accession number (eg 'HLA00001')
           1. Sequence version number (eg 'SV 1')
           2. ??? eg 'standard'
           3. Molecule type (e.g. 'DNA')
           4. Taxonomic division (e.g. 'HUM')
           5. Sequence length (e.g. '3503 BP.')
        """
        consumer.locus(fields[0])

        # See TODO on the EMBL _feed_first_line_new about version field
        version_parts = fields[1].split()
        if (
            len(version_parts) == 2
            and version_parts[0] == "SV"
            and version_parts[1].isdigit()
        ):
            consumer.version_suffix(version_parts[1])

        consumer.residue_type(fields[3])
        if "circular" in fields[3]:
            consumer.topology("circular")
            consumer.molecule_type(fields[3].replace("circular", "").strip())
        elif "linear" in fields[3]:
            consumer.topology("linear")
            consumer.molecule_type(fields[3].replace("linear", "").strip())
        else:
            consumer.molecule_type(fields[3].strip())
        consumer.data_file_division(fields[4])
        self._feed_seq_length(consumer, fields[5])

    def parse_features(self, skip=False):
        """Return list of tuples for the features (if present).

        Each feature is returned as a tuple (key, location, qualifiers)
        where key and location are strings (e.g. "CDS" and
        "complement(join(490883..490885,1..879))") while qualifiers
        is a list of two string tuples (feature qualifier keys and values).

        Assumes you have already read to the start of the features table.
        """
        if self.line.rstrip() not in self.FEATURE_START_MARKERS:
            if self.debug:
                print("Didn't find any feature table")
            return []

        while self.line.rstrip() in self.FEATURE_START_MARKERS:
            self.line = self.handle.readline()

        bad_position_re = re.compile(r"([0-9]+)>")

        features = []
        line = self.line
        while True:
            if not line:
                raise ValueError("Premature end of line during features table")
            if line[: self.HEADER_WIDTH].rstrip() in self.SEQUENCE_HEADERS:
                if self.debug:
                    print("Found start of sequence")
                break
            line = line.rstrip()
            if line == "//":
                raise ValueError("Premature end of features table, marker '//' found")
            if line in self.FEATURE_END_MARKERS:
                if self.debug:
                    print("Found end of features")
                line = self.handle.readline()
                break
            if line[2 : self.FEATURE_QUALIFIER_INDENT].strip() == "":
                # This is an empty feature line between qualifiers. Empty
                # feature lines within qualifiers are handled below (ignored).
                line = self.handle.readline()
                continue

            if skip:
                line = self.handle.readline()
                while (
                    line[: self.FEATURE_QUALIFIER_INDENT]
                    == self.FEATURE_QUALIFIER_SPACER
                ):
                    line = self.handle.readline()
            else:
                assert line[:2] == "FT"
                try:
                    feature_key, location_start = line[2:].strip().split()
                except ValueError:
                    # e.g. "FT   TRANSMEMBRANE-REGION2163..2240\n"
                    # Assume indent of 25 as per IMGT spec, with the location
                    # start in column 26 (one-based).
                    feature_key = line[2:25].strip()
                    location_start = line[25:].strip()
                feature_lines = [location_start]
                line = self.handle.readline()
                while (
                    line[: self.FEATURE_QUALIFIER_INDENT]
                    == self.FEATURE_QUALIFIER_SPACER
                    or line.rstrip() == ""
                ):  # cope with blank lines in the midst of a feature
                    # Use strip to remove any harmless trailing white space AND and leading
                    # white space (copes with 21 or 26 indents and orther variants)
                    assert line[:2] == "FT"
                    feature_lines.append(line[self.FEATURE_QUALIFIER_INDENT :].strip())
                    line = self.handle.readline()
                feature_key, location, qualifiers = self.parse_feature(
                    feature_key, feature_lines
                )
                # Try to handle known problems with IMGT locations here:
                if ">" in location:
                    # Nasty hack for common IMGT bug, should be >123 not 123>
                    # in a location string. At least here the meaning is clear,
                    # and since it is so common I don't want to issue a warning
                    # warnings.warn("Feature location %s is invalid, "
                    #              "moving greater than sign before position"
                    #              % location, BiopythonParserWarning)
                    location = bad_position_re.sub(r">\1", location)
                features.append((feature_key, location, qualifiers))
        self.line = line
        return features


class GenBankScanner(InsdcScanner):
    """For extracting chunks of information in GenBank files."""

    RECORD_START = "LOCUS       "
    HEADER_WIDTH = 12
    FEATURE_START_MARKERS = ["FEATURES             Location/Qualifiers", "FEATURES"]
    FEATURE_END_MARKERS = []
    FEATURE_QUALIFIER_INDENT = 21
    FEATURE_QUALIFIER_SPACER = " " * FEATURE_QUALIFIER_INDENT
    SEQUENCE_HEADERS = [
        "CONTIG",
        "ORIGIN",
        "BASE COUNT",
        "WGS",
        "TSA",
        "TLS",
    ]  # trailing spaces removed

    GENBANK_INDENT = HEADER_WIDTH
    GENBANK_SPACER = " " * GENBANK_INDENT

    STRUCTURED_COMMENT_START = "-START##"
    STRUCTURED_COMMENT_END = "-END##"
    STRUCTURED_COMMENT_DELIM = " :: "

    def parse_footer(self):
        """Return a tuple containing a list of any misc strings, and the sequence."""
        if self.line[: self.HEADER_WIDTH].rstrip() not in self.SEQUENCE_HEADERS:
            raise ValueError("Footer format unexpected:  '%s'" % self.line)

        misc_lines = []
        while (
            self.line[: self.HEADER_WIDTH].rstrip() in self.SEQUENCE_HEADERS
            or self.line[: self.HEADER_WIDTH] == " " * self.HEADER_WIDTH
            or "WGS" == self.line[:3]
        ):
            misc_lines.append(self.line.rstrip())
            self.line = self.handle.readline()
            if not self.line:
                raise ValueError("Premature end of file")

        if self.line[: self.HEADER_WIDTH].rstrip() in self.SEQUENCE_HEADERS:
            raise ValueError("Eh? '%s'" % self.line)

        # Now just consume the sequence lines until reach the // marker
        # or a CONTIG line
        seq_lines = []
        line = self.line
        while True:
            if not line:
                warnings.warn(
                    "Premature end of file in sequence data", BiopythonParserWarning
                )
                line = "//"
                break
            line = line.rstrip()
            if not line:
                warnings.warn("Blank line in sequence data", BiopythonParserWarning)
                line = self.handle.readline()
                continue
            if line == "//":
                break
            if line.startswith("CONTIG"):
                break
            if len(line) > 9 and line[9:10] != " ":
                # Some broken programs indent the sequence by one space too many
                # so try to get rid of that and test again.
                warnings.warn(
                    "Invalid indentation for sequence line", BiopythonParserWarning
                )
                line = line[1:]
                if len(line) > 9 and line[9:10] != " ":
                    raise ValueError("Sequence line mal-formed, '%s'" % line)
            seq_lines.append(line[10:])  # remove spaces later
            line = self.handle.readline()

        self.line = line
        return misc_lines, "".join(seq_lines).replace(" ", "")

    def _feed_first_line(self, consumer, line):
        """Scan over and parse GenBank LOCUS line (PRIVATE).

        This must cope with several variants, primarily the old and new column
        based standards from GenBank. Additionally EnsEMBL produces GenBank
        files where the LOCUS line is space separated rather that following
        the column based layout.

        We also try to cope with GenBank like files with partial LOCUS lines.

        As of release 229.0, the columns are no longer strictly in a given
        position. See GenBank format release notes:

            "Historically, the LOCUS line has had a fixed length and its
            elements have been presented at specific column positions...
            But with the anticipated increases in the lengths of accession
            numbers, and the advent of sequences that are gigabases long,
            maintaining the column positions will not always be possible and
            the overall length of the LOCUS line could exceed 79 characters."

        """
        #####################################
        # LOCUS line                        #
        #####################################
        if line[0 : self.GENBANK_INDENT] != "LOCUS       ":
            raise ValueError("LOCUS line does not start correctly:\n" + line)

        # Have to break up the locus line, and handle the different bits of it.
        # There are at least two different versions of the locus line...
        if line[29:33] in [" bp ", " aa ", " rc "] and line[55:62] == "       ":
            # Old... note we insist on the 55:62 being empty to avoid trying
            # to parse space separated LOCUS lines from Ensembl etc, see below.
            #
            #    Positions  Contents
            #    ---------  --------
            #    00:06      LOCUS
            #    06:12      spaces
            #    12:??      Locus name
            #    ??:??      space
            #    ??:29      Length of sequence, right-justified
            #    29:33      space, bp, space
            #    33:41      strand type / molecule type, e.g. DNA
            #    41:42      space
            #    42:51      Blank (implies linear), linear or circular
            #    51:52      space
            #    52:55      The division code (e.g. BCT, VRL, INV)
            #    55:62      space
            #    62:73      Date, in the form dd-MMM-yyyy (e.g., 15-MAR-1991)
            #
            # assert line[29:33] in [' bp ', ' aa ',' rc '] , \
            #       'LOCUS line does not contain size units at expected position:\n' + line
            if line[41:42] != " ":
                raise ValueError(
                    "LOCUS line does not contain space at position 42:\n" + line
                )
            if line[42:51].strip() not in ["", "linear", "circular"]:
                raise ValueError(
                    "LOCUS line does not contain valid entry "
                    "(linear, circular, ...):\n" + line
                )
            if line[51:52] != " ":
                raise ValueError(
                    "LOCUS line does not contain space at position 52:\n" + line
                )
            # if line[55:62] != '       ':
            #      raise ValueError('LOCUS line does not contain spaces from position 56 to 62:\n' + line)
            if line[62:73].strip():
                if line[64:65] != "-":
                    raise ValueError(
                        "LOCUS line does not contain - at "
                        "position 65 in date:\n" + line
                    )
                if line[68:69] != "-":
                    raise ValueError(
                        "LOCUS line does not contain - at "
                        "position 69 in date:\n" + line
                    )

            name_and_length_str = line[self.GENBANK_INDENT : 29]
            while "  " in name_and_length_str:
                name_and_length_str = name_and_length_str.replace("  ", " ")
            name_and_length = name_and_length_str.split(" ")
            if len(name_and_length) > 2:
                raise ValueError(
                    "Cannot parse the name and length in the LOCUS line:\n" + line
                )
            if len(name_and_length) == 1:
                raise ValueError("Name and length collide in the LOCUS line:\n" + line)
            # Should be possible to split them based on position, if
            # a clear definition of the standard exists THAT AGREES with
            # existing files.
            name, length = name_and_length
            if len(name) > 16:
                # As long as the sequence is short, can steal its leading spaces
                # to extend the name over the current 16 character limit.
                # However, that deserves a warning as it is out of spec.
                warnings.warn(
                    "GenBank LOCUS line identifier over 16 characters",
                    BiopythonParserWarning,
                )
            consumer.locus(name)
            consumer.size(length)
            # consumer.residue_type(line[33:41].strip())

            if line[33:51].strip() == "" and line[29:33] == " aa ":
                # Amino acids -> protein (even if there is no residue type given)
                consumer.residue_type("PROTEIN")
            else:
                consumer.residue_type(line[33:51].strip())

            consumer.molecule_type(line[33:41].strip())
            consumer.topology(line[42:51].strip())
            consumer.data_file_division(line[52:55])
            if line[62:73].strip():
                consumer.date(line[62:73])
        elif line[40:44] in [" bp ", " aa ", " rc "] and line[54:64].strip() in [
            "",
            "linear",
            "circular",
        ]:
            # New... linear/circular/big blank test should avoid EnsEMBL style
            # LOCUS line being treated like a proper column based LOCUS line.
            #
            #    Positions  Contents
            #    ---------  --------
            #    00:06      LOCUS
            #    06:12      spaces
            #    12:??      Locus name
            #    ??:??      space
            #    ??:40      Length of sequence, right-justified
            #    40:44      space, bp, space
            #    44:47      Blank, ss-, ds-, ms-
            #    47:54      Blank, DNA, RNA, tRNA, mRNA, uRNA, snRNA, cDNA
            #    54:55      space
            #    55:63      Blank (implies linear), linear or circular
            #    63:64      space
            #    64:67      The division code (e.g. BCT, VRL, INV)
            #    67:68      space
            #    68:79      Date, in the form dd-MMM-yyyy (e.g., 15-MAR-1991)
            #
            if len(line) < 79:
                # JBEI genbank files seem to miss a division code and date
                # See issue #1656 e.g.
                # LOCUS       pEH010                  5743 bp    DNA     circular
                warnings.warn(
                    "Truncated LOCUS line found - is this correct?\n:%r" % line,
                    BiopythonParserWarning,
                )
                padding_len = 79 - len(line)
                padding = " " * padding_len
                line += padding

            if line[40:44] not in [" bp ", " aa ", " rc "]:
                raise ValueError(
                    "LOCUS line does not contain size units at "
                    "expected position:\n" + line
                )
            if line[44:47] not in ["   ", "ss-", "ds-", "ms-"]:
                raise ValueError(
                    "LOCUS line does not have valid strand "
                    "type (Single stranded, ...):\n" + line
                )

            if not (
                line[47:54].strip() == ""
                or "DNA" in line[47:54].strip().upper()
                or "RNA" in line[47:54].strip().upper()
            ):
                raise ValueError(
                    "LOCUS line does not contain valid "
                    "sequence type (DNA, RNA, ...):\n" + line
                )
            if line[54:55] != " ":
                raise ValueError(
                    "LOCUS line does not contain space at position 55:\n" + line
                )
            if line[55:63].strip() not in ["", "linear", "circular"]:
                raise ValueError(
                    "LOCUS line does not contain valid "
                    "entry (linear, circular, ...):\n" + line
                )
            if line[63:64] != " ":
                raise ValueError(
                    "LOCUS line does not contain space at position 64:\n" + line
                )
            if line[67:68] != " ":
                raise ValueError(
                    "LOCUS line does not contain space at position 68:\n" + line
                )
            if line[68:79].strip():
                if line[70:71] != "-":
                    raise ValueError(
                        "LOCUS line does not contain - at "
                        "position 71 in date:\n" + line
                    )
                if line[74:75] != "-":
                    raise ValueError(
                        "LOCUS line does not contain - at "
                        "position 75 in date:\n" + line
                    )

            name_and_length_str = line[self.GENBANK_INDENT : 40]
            while "  " in name_and_length_str:
                name_and_length_str = name_and_length_str.replace("  ", " ")
            name_and_length = name_and_length_str.split(" ")
            if len(name_and_length) > 2:
                raise ValueError(
                    "Cannot parse the name and length in the LOCUS line:\n" + line
                )
            if len(name_and_length) == 1:
                raise ValueError("Name and length collide in the LOCUS line:\n" + line)
            # Should be possible to split them based on position, if
            # a clear definition of the stand exists THAT AGREES with
            # existing files.
            consumer.locus(name_and_length[0])
            consumer.size(name_and_length[1])

            if line[44:54].strip() == "" and line[40:44] == " aa ":
                # Amino acids -> protein (even if there is no residue type given)
                consumer.residue_type(("PROTEIN " + line[54:63]).strip())
            else:
                consumer.residue_type(line[44:63].strip())

            consumer.molecule_type(line[44:54].strip())
            consumer.topology(line[55:63].strip())
            if line[64:76].strip():
                consumer.data_file_division(line[64:67])
            if line[68:79].strip():
                consumer.date(line[68:79])
        elif line[self.GENBANK_INDENT :].strip().count(" ") == 0:
            # Truncated LOCUS line, as produced by some EMBOSS tools - see bug 1762
            #
            # e.g.
            #
            #    "LOCUS       U00096"
            #
            # rather than:
            #
            #    "LOCUS       U00096               4639675 bp    DNA     circular BCT"
            #
            #    Positions  Contents
            #    ---------  --------
            #    00:06      LOCUS
            #    06:12      spaces
            #    12:??      Locus name
            if line[self.GENBANK_INDENT :].strip() != "":
                consumer.locus(line[self.GENBANK_INDENT :].strip())
            else:
                # Must just have just "LOCUS       ", is this even legitimate?
                # We should be able to continue parsing... we need real world testcases!
                warnings.warn(
                    "Minimal LOCUS line found - is this correct?\n:%r" % line,
                    BiopythonParserWarning,
                )
        elif (
            len(line.split()) == 8
            and line.split()[3] in ("aa", "bp")
            and line.split()[5] in ("linear", "circular")
        ):
            # Cope with invalidly spaced GenBank LOCUS lines like
            # LOCUS       AB070938          6497 bp    DNA     linear   BCT 11-OCT-2001
            # This will also cope with extra long accession numbers and
            # sequence lengths
            splitline = line.split()
            consumer.locus(splitline[1])
            # Provide descriptive error message if the sequence is too long
            # for python to handle

            if int(splitline[2]) > sys.maxsize:
                raise ValueError(
                    "Tried to load a sequence with a length %s, "
                    "your installation of python can only load "
                    "sesquences of length %s" % (splitline[2], sys.maxsize)
                )
            else:
                consumer.size(splitline[2])

            consumer.residue_type(splitline[4])
            consumer.topology(splitline[5])
            consumer.data_file_division(splitline[6])
            consumer.date(splitline[7])
            if len(line) < 80:
                warnings.warn(
                    "Attempting to parse malformed locus line:\n%r\n"
                    "Found locus %r size %r residue_type %r\n"
                    "Some fields may be wrong."
                    % (line, splitline[1], splitline[2], splitline[4]),
                    BiopythonParserWarning,
                )
        elif len(line.split()) == 7 and line.split()[3] in ["aa", "bp"]:
            # Cope with EnsEMBL genbank files which use space separation rather
            # than the expected column based layout. e.g.
            # LOCUS       HG531_PATCH 1000000 bp DNA HTG 18-JUN-2011
            # LOCUS       HG531_PATCH 759984 bp DNA HTG 18-JUN-2011
            # LOCUS       HG506_HG1000_1_PATCH 814959 bp DNA HTG 18-JUN-2011
            # LOCUS       HG506_HG1000_1_PATCH 1219964 bp DNA HTG 18-JUN-2011
            # Notice that the 'bp' can occur in the position expected by either
            # the old or the new fixed column standards (parsed above).
            splitline = line.split()
            consumer.locus(splitline[1])
            consumer.size(splitline[2])
            consumer.residue_type(splitline[4])
            consumer.data_file_division(splitline[5])
            consumer.date(splitline[6])
        elif len(line.split()) >= 4 and line.split()[3] in ["aa", "bp"]:
            # Cope with EMBOSS seqret output where it seems the locus id can cause
            # the other fields to overflow.  We just IGNORE the other fields!
            warnings.warn(
                "Malformed LOCUS line found - is this correct?\n:%r" % line,
                BiopythonParserWarning,
            )
            consumer.locus(line.split()[1])
            consumer.size(line.split()[2])
        elif len(line.split()) >= 4 and line.split()[-1] in ["aa", "bp"]:
            # Cope with pseudo-GenBank files like this:
            #   "LOCUS       RNA5 complete       1718 bp"
            # Treat everything between LOCUS and the size as the identifier.
            warnings.warn(
                "Malformed LOCUS line found - is this correct?\n:%r" % line,
                BiopythonParserWarning,
            )
            consumer.locus(line[5:].rsplit(None, 2)[0].strip())
            consumer.size(line.split()[-2])
        else:
            raise ValueError("Did not recognise the LOCUS line layout:\n" + line)

    def _feed_header_lines(self, consumer, lines):
        # Following dictionary maps GenBank lines to the associated
        # consumer methods - the special cases like LOCUS where one
        # genbank line triggers several consumer calls have to be
        # handled individually.
        consumer_dict = {
            "DEFINITION": "definition",
            "ACCESSION": "accession",
            "NID": "nid",
            "PID": "pid",
            "DBSOURCE": "db_source",
            "KEYWORDS": "keywords",
            "SEGMENT": "segment",
            "SOURCE": "source",
            "AUTHORS": "authors",
            "CONSRTM": "consrtm",
            "PROJECT": "project",
            "TITLE": "title",
            "JOURNAL": "journal",
            "MEDLINE": "medline_id",
            "PUBMED": "pubmed_id",
            "REMARK": "remark",
        }
        # We have to handle the following specially:
        # ORIGIN (locus, size, residue_type, data_file_division and date)
        # COMMENT (comment)
        # VERSION (version and gi)
        # DBLINK (database links like projects, newlines important)
        # REFERENCE (eference_num and reference_bases)
        # ORGANISM (organism and taxonomy)
        lines = [_f for _f in lines if _f]
        lines.append("")  # helps avoid getting StopIteration all the time
        line_iter = iter(lines)
        try:
            line = next(line_iter)
            while True:
                if not line:
                    break
                line_type = line[: self.GENBANK_INDENT].strip()
                data = line[self.GENBANK_INDENT :].strip()

                if line_type == "VERSION":
                    # Need to call consumer.version(), and maybe also consumer.gi() as well.
                    # e.g.
                    # VERSION     AC007323.5  GI:6587720
                    while "  " in data:
                        data = data.replace("  ", " ")
                    if " GI:" not in data:
                        consumer.version(data)
                    else:
                        if self.debug:
                            print(
                                "Version ["
                                + data.split(" GI:")[0]
                                + "], gi ["
                                + data.split(" GI:")[1]
                                + "]"
                            )
                        consumer.version(data.split(" GI:")[0])
                        consumer.gi(data.split(" GI:")[1])
                    # Read in the next line!
                    line = next(line_iter)
                elif line_type == "DBLINK":
                    # Need to call consumer.dblink() for each line, e.g.
                    # DBLINK      Project: 57779
                    #             BioProject: PRJNA57779
                    consumer.dblink(data.strip())
                    # Read in the next line, and see if its more of the DBLINK section:
                    while True:
                        line = next(line_iter)
                        if line[: self.GENBANK_INDENT] == self.GENBANK_SPACER:
                            # Add this continuation to the data string
                            consumer.dblink(line[self.GENBANK_INDENT :].strip())
                        else:
                            # End of the DBLINK, leave this text in the variable "line"
                            break
                elif line_type == "REFERENCE":
                    if self.debug > 1:
                        print("Found reference [" + data + "]")
                    # Need to call consumer.reference_num() and consumer.reference_bases()
                    # e.g.
                    # REFERENCE   1  (bases 1 to 86436)
                    #
                    # Note that this can be multiline, see Bug 1968, e.g.
                    #
                    # REFERENCE   42 (bases 1517 to 1696; 3932 to 4112; 17880 to 17975; 21142 to
                    #             28259)
                    #
                    # For such cases we will call the consumer once only.
                    data = data.strip()

                    # Read in the next line, and see if its more of the reference:
                    while True:
                        line = next(line_iter)
                        if line[: self.GENBANK_INDENT] == self.GENBANK_SPACER:
                            # Add this continuation to the data string
                            data += " " + line[self.GENBANK_INDENT :]
                            if self.debug > 1:
                                print("Extended reference text [" + data + "]")
                        else:
                            # End of the reference, leave this text in the variable "line"
                            break

                    # We now have all the reference line(s) stored in a string, data,
                    # which we pass to the consumer
                    while "  " in data:
                        data = data.replace("  ", " ")
                    if " " not in data:
                        if self.debug > 2:
                            print('Reference number "' + data + '"')
                        consumer.reference_num(data)
                    else:
                        if self.debug > 2:
                            print(
                                'Reference number "'
                                + data[: data.find(" ")]
                                + '", "'
                                + data[data.find(" ") + 1 :]
                                + '"'
                            )
                        consumer.reference_num(data[: data.find(" ")])
                        consumer.reference_bases(data[data.find(" ") + 1 :])
                elif line_type == "ORGANISM":
                    # Typically the first line is the organism, and subsequent lines
                    # are the taxonomy lineage.  However, given longer and longer
                    # species names (as more and more strains and sub strains get
                    # sequenced) the oragnism name can now get wrapped onto multiple
                    # lines.  The NCBI say we have to recognise the lineage line by
                    # the presence of semi-colon delimited entries.  In the long term,
                    # they are considering adding a new keyword (e.g. LINEAGE).
                    # See Bug 2591 for details.
                    organism_data = data
                    lineage_data = ""
                    while True:
                        line = next(line_iter)
                        if line[0 : self.GENBANK_INDENT] == self.GENBANK_SPACER:
                            if lineage_data or ";" in line:
                                lineage_data += " " + line[self.GENBANK_INDENT :]
                            elif line[self.GENBANK_INDENT :].strip() == ".":
                                # No lineage data, just . place holder
                                pass
                            else:
                                organism_data += (
                                    " " + line[self.GENBANK_INDENT :].strip()
                                )
                        else:
                            # End of organism and taxonomy
                            break
                    consumer.organism(organism_data)
                    if lineage_data.strip() == "" and self.debug > 1:
                        print("Taxonomy line(s) missing or blank")
                    consumer.taxonomy(lineage_data.strip())
                    del organism_data, lineage_data
                elif line_type == "COMMENT":
                    # A COMMENT can either be plain text or tabular (Structured Comment),
                    # or contain both. Multi-line comments are common. The code calls
                    # consumer.comment() once with a list where each entry
                    # is a line. If there's a structured comment consumer.structured_comment()
                    # is called with a dict of dicts where the secondary key/value pairs are
                    # the same as those in the structured comment table. The primary key is
                    # the title or header of the table (e.g. Assembly-Data, FluData). See
                    # http://www.ncbi.nlm.nih.gov/genbank/structuredcomment
                    # for more information on Structured Comments.
                    data = line[self.GENBANK_INDENT :]
                    if self.debug > 1:
                        print("Found comment")
                    comment_list = []
                    structured_comment_dict = defaultdict(dict)
                    regex = fr"([^#]+){self.STRUCTURED_COMMENT_START}$"
                    structured_comment_key = re.search(regex, data)
                    if structured_comment_key is not None:
                        structured_comment_key = structured_comment_key.group(1)
                        if self.debug > 1:
                            print("Found Structured Comment")
                    else:
                        comment_list.append(data)

                    while True:
                        line = next(line_iter)
                        data = line[self.GENBANK_INDENT :]
                        if line[0 : self.GENBANK_INDENT] == self.GENBANK_SPACER:
                            if self.STRUCTURED_COMMENT_START in data:
                                regex = r"([^#]+){}$".format(
                                    self.STRUCTURED_COMMENT_START
                                )
                                structured_comment_key = re.search(regex, data)
                                if structured_comment_key is not None:
                                    structured_comment_key = structured_comment_key.group(
                                        1
                                    )
                                else:
                                    comment_list.append(data)
                            elif (
                                structured_comment_key is not None
                                and self.STRUCTURED_COMMENT_DELIM in data
                            ):
                                match = re.search(
                                    r"(.+?)\s*{}\s*(.+)".format(
                                        self.STRUCTURED_COMMENT_DELIM
                                    ),
                                    data,
                                )
                                structured_comment_dict[structured_comment_key][
                                    match.group(1)
                                ] = match.group(2)
                                if self.debug > 2:
                                    print(
                                        "Structured Comment continuation [" + data + "]"
                                    )
                            elif (
                                structured_comment_key is not None
                                and self.STRUCTURED_COMMENT_END not in data
                            ):
                                # Don't die on a malformed comment, just warn and carry on
                                if (
                                    structured_comment_key
                                    not in structured_comment_dict
                                ):
                                    warnings.warn(
                                        "Structured comment not parsed for %s. Is it malformed?"
                                        % consumer.data.name,
                                        BiopythonParserWarning,
                                    )
                                    continue

                                # The current structured comment has a multiline value
                                previous_value_line = structured_comment_dict[
                                    structured_comment_key
                                ][match.group(1)]
                                structured_comment_dict[structured_comment_key][
                                    match.group(1)
                                ] = (previous_value_line + " " + line.strip())
                            elif self.STRUCTURED_COMMENT_END in data:
                                # End of structured comment
                                structured_comment_key = None
                            else:
                                comment_list.append(data)
                                if self.debug > 2:
                                    print("Comment continuation [" + data + "]")
                        else:
                            # End of the comment
                            break
                    if comment_list:
                        consumer.comment(comment_list)
                    if structured_comment_dict:
                        consumer.structured_comment(structured_comment_dict)
                    del comment_list, structured_comment_key, structured_comment_dict
                elif line_type in consumer_dict:
                    # It's a semi-automatic entry!
                    # Now, this may be a multi line entry...
                    while True:
                        line = next(line_iter)
                        if line[0 : self.GENBANK_INDENT] == self.GENBANK_SPACER:
                            data += " " + line[self.GENBANK_INDENT :]
                        else:
                            # We now have all the data for this entry:

                            # The DEFINITION field must ends with a period
                            # # see ftp://ftp.ncbi.nih.gov/genbank/gbrel.txt [3.4.5]
                            # and discussion https://github.com/biopython/biopython/pull/616
                            # We consider this period belong to the syntax, not to the data
                            # So remove it if it exist
                            if line_type == "DEFINITION" and data.endswith("."):
                                data = data[:-1]
                            getattr(consumer, consumer_dict[line_type])(data)
                            # End of continuation - return to top of loop!
                            break
                else:
                    if self.debug:
                        print("Ignoring GenBank header line:\n" % line)
                    # Read in next line
                    line = next(line_iter)
        except StopIteration:
            raise ValueError("Problem in header") from None

    def _feed_misc_lines(self, consumer, lines):
        # Deals with a few misc lines between the features and the sequence
        lines.append("")
        line_iter = iter(lines)
        try:
            for line in line_iter:
                if line.startswith("BASE COUNT"):
                    line = line[10:].strip()
                    if line:
                        if self.debug:
                            print("base_count = " + line)
                        consumer.base_count(line)
                if line.startswith("ORIGIN"):
                    line = line[6:].strip()
                    if line:
                        if self.debug:
                            print("origin_name = " + line)
                        consumer.origin_name(line)
                if line.startswith("TLS "):
                    line = line[3:].strip()
                    consumer.tls(line)
                if line.startswith("TSA "):
                    line = line[3:].strip()
                    consumer.tsa(line)
                if line.startswith("WGS "):
                    line = line[3:].strip()
                    consumer.wgs(line)
                if line.startswith("WGS_SCAFLD"):
                    line = line[10:].strip()
                    consumer.add_wgs_scafld(line)
                if line.startswith("CONTIG"):
                    line = line[6:].strip()
                    contig_location = line
                    while True:
                        line = next(line_iter)
                        if not line:
                            break
                        elif line[: self.GENBANK_INDENT] == self.GENBANK_SPACER:
                            # Don't need to preseve the whitespace here.
                            contig_location += line[self.GENBANK_INDENT :].rstrip()
                        elif line.startswith("ORIGIN"):
                            # Strange, seen this in GenPept files via Entrez gbwithparts
                            line = line[6:].strip()
                            if line:
                                consumer.origin_name(line)
                            break
                        else:
                            raise ValueError(
                                "Expected CONTIG continuation line, got:\n" + line
                            )
                    consumer.contig_location(contig_location)
            return
        except StopIteration:
            raise ValueError("Problem in misc lines before sequence") from None