Implement a less aggressive mode for heuristic dating supplementation.

Dating supplementation works on the assumption that when a source states
a witness has not been created after a certain date but says nothing
about the earliest possible moment it still suggests a certain closeness
of the creation date to that terminus ante quem (and the same is valid
the other way around).

The behaviour can be configured using the new option half_interval_mode:
- always: whenever a source gives only one end of the possible dating
  interval, supplement the other interval limit
- light: when we have only sources about one interval limit but none
  about the other, take the strongest source and infer the other interval
  from that.

The length of the artificial interval can be configured by half_interval_correction.

src/macrogen/bibliography.py

@@ -14,6 +14,14 @@
 class BiblSource:
     """
     A bibliographic source in a macrogenesis XML file.
+
+    Attributes and properties:
+        uri (str): The faust://bibliography URI
+        detail (str): Detail string like pages
+        weight (int): Score for the source
+        filename (str): Representation of the source (w/o detail) that is usable as part of a filename
+        citation (str): Short citation
+        long_citation (str): Detail string for the citation
     """
 
     def __init__(self, uri, detail=''):

src/macrogen/datings.py

@@ -173,19 +173,19 @@ def add_to_graph(self, G: nx.MultiDiGraph):
                 if self.start is not None:
                     G.add_edge(self.date_before, item, kind=self.start_attr[0], source=source, dating=self,
                                xml=self.xmlsource, ignore=self.ignore, comments=self.comments)
-                    if self.end is None and not self.ignore:
+                    if self.end is None and config.half_interval_mode == 'always' and not self.ignore:
                         G.add_edge(item, min(self.date_before + timedelta(config.half_interval_correction),
                                              date(1832, 3, 23)), kind='not_after',
                                    trigger_node=self.date_before,
-                                   source=BiblSource('faust://heuristic'), xml=self.xmlsource)
+                                   source=BiblSource('faust://heuristic', source.citation), xml=self.xmlsource)
                 if self.end is not None:
                     G.add_edge(item, self.date_after, kind=self.end_attr[0], source=source, dating=self,
                                xml=self.xmlsource, ignore=self.ignore, comments=self.comments)
-                    if self.start is None and not self.ignore:
+                    if self.start is None and config.half_interval_mode == 'always' and not self.ignore:
                         G.add_edge(self.date_after - timedelta(config.half_interval_correction), item,
                                    kind='not_before',
                                    trigger_node=self.date_after,
-                                   source=BiblSource('faust://heuristic'), xml=self.xmlsource)
+                                   source=BiblSource('faust://heuristic', source.citation), xml=self.xmlsource)
 
 
 class RelativeDating(_AbstractDating):
@@ -268,6 +268,23 @@ def build_datings_graph() -> nx.MultiDiGraph:
     logger.info('Reading data to build base graph ...')
     for dating in _parse_files():
         dating.add_to_graph(graph)
+
+    if config.half_interval_mode == 'light':
+        for node in list(graph.nodes):
+            if isinstance(node, Reference):
+                post = [(v, attr) for (_, v, attr) in graph.out_edges(node, data=True) if isinstance(v, date)]
+                pre = [(u, attr) for (u, _, attr) in graph.in_edges(node, data=True) if isinstance(u, date)]
+                if pre and not post:
+                    other_limit = max(pre, key=lambda item: item[1]['source'].weight)
+                    graph.add_edge(node, other_limit[0] + timedelta(config.half_interval_correction),
+                                   source=BiblSource('faust://heuristic', other_limit[1]['source'].citation),
+                                   kind='not_after', xml=other_limit[1]['xml'])
+                elif post and not pre:
+                    other_limit = max(post, key=lambda item: item[1]['source'].weight)
+                    graph.add_edge(other_limit[0] - timedelta(config.half_interval_correction), node,
+                                   source=BiblSource('faust://heuristic', other_limit[1]['source'].citation),
+                                   kind='not_before', xml=other_limit[1]['xml'])
+
     add_timeline_edges(graph)
     return graph
 

src/macrogen/etc/default.yaml

@@ -19,6 +19,7 @@ bibliography: https://raw.githubusercontent.com/faustedition/faust-gen-html/mast
 xmlroot: https://github.com/faustedition/faust-xml/tree/master/xml/macrogenesis  # base for links to xml files
 
 ## Limits
+half_interval_mode: light          # off, light, always
 half_interval_correction: 182.5    # if we only have a start or end date, the other limit is max. this many days away
 render_node_limit: 1500            # do not layout graphs with more nodes than this
 render_timeout:                    # max number of seconds before rendering a dot file is aborted