Add the authors modules containing a collection of iterators.

modules/invenio_tools/__init__.py

@@ -26,6 +26,12 @@ from exception import (CdsException,
                        RecordException,
                        XmlException)
 
+from authors import (iter_author_affiliations,
+                     iter_author_affiliation_keys,
+                     iter_author_fields,
+                     iter_author_items,
+                     iter_author_names)
+
 from inveniostore import InvenioStore
 from iterrecord import IterRecord, REG_INT
 from marc12 import Marc12

modules/invenio_tools/authors.py

+# -*- coding: utf-8 -*-
+""" invenio_tools.authors
+
+"""
+from base import REG_AUTHOR
+from itertools import chain, imap, izip_longest
+
+to_list = lambda x: (x if isinstance(x, list) else [x])
+
+
+def affiliation_keys(field):
+    """Extract affiliation key(s) from the author field.
+
+    Note:
+        The affiliation keys are obtained by concatenating the "u" and "v"
+        keys. Keep in mind that an author can have several affiliations.
+
+    Args:
+        field (dict): author field
+
+    Returns:
+        itertools.imap
+
+    """
+    if "u" not in field:
+        return ()
+
+    key_u, key_v = to_list(field["u"]), []
+    if "v" in field:
+        key_v = to_list(field["v"])
+
+    # tuple with u and v key
+    iter_uv = izip_longest(key_u, key_v, fillvalue="")
+
+    # concatenate u and v key
+    return imap(lambda x: u"%s%s" % x, iter_uv)
+
+
+def author_item(field):
+    """Extract the item from the author field.
+
+    The author item is a tuple containing the author name and an iterator
+    on the author affiliation keys.
+
+    Args:
+        field (dict): author field
+
+    Returns:
+        tuple: (author name, affiliation keys iterator)
+
+    """
+    if "a" not in field:
+        return (u"", u"")
+
+    return (field["a"], affiliation_keys(field))
+
+
+def author_name(field):
+    """Extract the name from the author field.
+
+    Note:
+        In most of the case the author is a string
+        but it can be a list, e.g inspirehep.net/138663:
+        [u'Zuniga, J.', u'(the A.N.T.ARES. Collaboration)']
+
+    Args:
+        field (dict): author field
+
+    Returns:
+        str:
+
+    """
+    name =  (field["a"] if "a" in field else u"")
+
+    if isinstance(name, list):
+        for elt in name:
+            if REG_AUTHOR.match(elt):
+                return elt
+
+    return name
+
+
+def iter_author_fields(record):
+    """Iterator over the author field.
+
+    The author are spread over the 100 and 700 field. The function takes into
+    cases where the first author is defined in 100 but not in 700, first author
+    is defined in 100 and in 700 or no author in 100.
+
+    Args:
+        record (Record): MARC record for the publication
+
+    Returns:
+        iterator: either a itertools.chain or a list. The list is empty
+            when the author field does not exist.
+
+    """
+    if u"100" in record and u"700" in record:
+
+        u100, u700 = record[u"100"], record[u"700"]
+
+        if isinstance(u100, dict) and isinstance(u700, list):
+            if "a" in u100 and "a" in u700[0] and u100["a"] == u700[0]["a"]:
+                return to_list(u700)
+
+        elif isinstance(u100, dict) and isinstance(u700, dict):
+            if "a" in u100 and "a" in u700 and u100["a"] == u700["a"]:
+                return to_list(u100)
+
+        return chain(to_list(u100), to_list(u700))
+
+    elif u"100" in record:
+        authors = to_list(record[u"100"])
+
+    elif u"700" in record:
+        authors = record[u"700"]
+
+    else:
+        return ()
+
+
+def iter_author_affiliations(record):
+    """Iterator over the author affiliations.
+
+    Note:
+        The affiliation is defined by the "u" key.
+        Keep in mind that an author can have several affiliations.
+
+    Args:
+        record (Record): MARC record for the publication
+
+    Returns:
+        itertools.imap
+
+    """
+    ftpl = lambda x: (tuple(x) if isinstance(x, (list, tuple)) else (x,))
+    func = lambda x: (ftpl(x["u"]) if "u" in x else ())
+    return imap(func, iter_author_fields(record))
+
+
+def iter_author_affiliation_keys(record):
+    """Iterator over the author affiliation keys.
+
+    Note:
+        The affiliation keys are obtained by concatenating the "u" and "v"
+        keys. Keep in mind that an author can have several affiliations.
+
+    Args:
+        record (Record): MARC record for the publication
+
+    Returns:
+        itertools.imap
+
+    """
+    return imap(affiliation_keys, iter_author_fields(record))
+
+
+def iter_author_items(record):
+    """Iterator over the author items.
+
+    The author item is a tuple containing the author name and an iterator
+    on the author affiliation keys.
+
+    Args:
+        record (Record): MARC record for the publication
+
+    Returns:
+        itertools.imap
+
+    """
+    return imap(author_item, iter_author_fields(record))
+
+
+def iter_author_names(record):
+    """Iterator over the author names.
+
+    Args:
+        record (Record): MARC record for the publication
+
+    Returns:
+        itertools.imap
+
+    """
+    return imap(author_name, iter_author_fields(record))

tests/invenio_tools/author_iterators/test_acl_cds1951625.py

+# -*- coding: utf-8 -*-
+"""ARTICLE
+
+    http://cds.cern.ch/record/1951625.
+
+    Precision luminosity measurements at LHCb,
+    J. Instrum. 9 (2014) P12005
+    arXiv:1410.0149
+    704 authors
+
+Note:
+    * The first author is not in the author list
+    * LHCb collaboration
+    * The publication year is a list (duplicate 773y)
+    * The submitted date is not formatted: 01 Oct 2014
+
+"""
+import itertools
+import pytest
+
+from invenio_tools import (iter_author_affiliation_keys,
+                           iter_author_fields,
+                           iter_author_items,
+                           iter_author_names,
+                           load_record)
+
+
+@pytest.fixture(scope="module")
+def record():
+    return load_record('cds.cern.ch', 1951625)
+
+
+def test_iter_author_fields(record):
+    assert type(iter_author_fields(record)) == itertools.chain
+
+
+def test_iter_names(record):
+    assert type(iter_author_names(record)) == itertools.imap
+
+
+def test_iter_affiliation_keys(record):
+    assert type(iter_author_affiliation_keys(record)) == itertools.imap
+
+
+def test_names(record):
+    names = list(iter_author_names(record))
+
+    assert len(names) == 704
+    assert names[0] == "Aaij, Roel"
+    assert names[1] == "Adeva, Bernardo"
+    assert names[344] == "Le Gac, Renaud"
+    assert names[-1] == "Zvyagin, Alexander"
+
+def test_affiliations(record):
+    affiliations = [tuple(el) for el in iter_author_affiliation_keys(record)]
+
+    assert len(affiliations) == 704
+    assert affiliations[0] == ("NIKHEF, Amsterdam",)
+    assert affiliations[-8] == (u"Heidelberg, Max Planck Inst.", u"Lebedev Inst.")
+
+
+def test_items(record):
+    items = [(el[0], tuple(el[1])) for el in iter_author_items(record)]
+
+    import pprint
+    pprint.pprint(items)
+
+    assert len(items) == 704
+    assert items[0] == (u'Aaij, Roel', (u'NIKHEF, Amsterdam',))
+    assert items[-8] == (u'Zavertyaev, Mikhail', (u'Heidelberg, Max Planck Inst.', u'Lebedev Inst.'))