Merge pull request #44 from eigenmagic/allowlist-thresholds

Config options for blocklist thresholds

src/fediblockhole/__init__.py

@@ -71,7 +71,7 @@ def sync_blocklists(conf: argparse.Namespace):
             import_fields, conf.save_intermediate, conf.savedir, export_fields))
 
     # Merge blocklists into an update dict
-    merged = merge_blocklists(blocklists, conf.mergeplan)
+    merged = merge_blocklists(blocklists, conf.mergeplan, conf.merge_threshold, conf.merge_threshold_type)
 
     # Remove items listed in allowlists, if any
     allowlists = fetch_allowlists(conf)
@@ -222,17 +222,21 @@ def merge_blocklists(blocklists: list[Blocklist], mergeplan: str='max',
         if threshold_type == 'count':
             domain_threshold_level = len(domain_blocks[domain])
         elif threshold_type == 'pct':
-            domain_threshold_level = len(domain_blocks[domain]) / num_blocklists
+            domain_threshold_level = len(domain_blocks[domain]) / num_blocklists * 100
+            # log.debug(f"domain threshold level: {domain_threshold_level}")
         else:
             raise ValueError(f"Unsupported threshold type '{threshold_type}'. Supported values are: 'count', 'pct'")
 
+        log.debug(f"Checking if {domain_threshold_level} >= {threshold} for {domain}")
         if domain_threshold_level >= threshold:
             # Add first block in the list to merged
             block = domain_blocks[domain][0]
+            log.debug(f"Yes. Merging block: {block}")
+
             # Merge the others with this record
             for newblock in domain_blocks[domain][1:]:
                 block = apply_mergeplan(block, newblock, mergeplan)
-        merged.blocks[block.domain] = block
+            merged.blocks[block.domain] = block
 
     return merged
 
@@ -706,6 +710,12 @@ def augment_args(args, tomldata: str=None):
     if not args.mergeplan:
         args.mergeplan = conf.get('mergeplan', 'max')
 
+    if not args.merge_threshold:
+        args.merge_threshold = conf.get('merge_threshold', 0)
+
+    if not args.merge_threshold_type:
+        args.merge_threshold_type = conf.get('merge_threshold_type', 'count')
+
     args.blocklist_url_sources = conf.get('blocklist_url_sources', [])
     args.blocklist_instance_sources = conf.get('blocklist_instance_sources', [])
     args.allowlist_url_sources = conf.get('allowlist_url_sources', [])
@@ -727,6 +737,8 @@ def setup_argparse():
     ap.add_argument('-S', '--save-intermediate', dest="save_intermediate", action='store_true', help="Save intermediate blocklists we fetch to local files.")
     ap.add_argument('-D', '--savedir', dest="savedir", help="Directory path to save intermediate lists.")
     ap.add_argument('-m', '--mergeplan', choices=['min', 'max'], help="Set mergeplan.")
+    ap.add_argument('--merge-threshold', type=int, help="Merge threshold value")
+    ap.add_argument('--merge-threshold-type', choices=['count', 'pct'], help="Type of merge threshold to use.")
 
     ap.add_argument('-I', '--import-field', dest='import_fields', action='append', help="Extra blocklist fields to import.")
     ap.add_argument('-E', '--export-field', dest='export_fields', action='append', help="Extra blocklist fields to export.")

tests/test_configfile.py

@@ -49,3 +49,33 @@ allowlist_url_sources = [ { url='file:///path/to/allowlist', format='csv'} ]
         'url': 'file:///path/to/allowlist',
         'format': 'csv',
         }]
+
+def test_set_merge_thresold_default():
+    tomldata = """
+"""
+    args = shim_argparse([], tomldata)
+
+    assert args.mergeplan == 'max'
+    assert args.merge_threshold_type == 'count'
+
+def test_set_merge_thresold_count():
+    tomldata = """# Add a merge threshold
+merge_threshold_type = 'count'
+merge_threshold = 2
+"""
+    args = shim_argparse([], tomldata)
+
+    assert args.mergeplan == 'max'
+    assert args.merge_threshold_type == 'count'
+    assert args.merge_threshold == 2
+
+def test_set_merge_thresold_pct():
+    tomldata = """# Add a merge threshold
+merge_threshold_type = 'pct'
+merge_threshold = 35
+"""
+    args = shim_argparse([], tomldata)
+
+    assert args.mergeplan == 'max'
+    assert args.merge_threshold_type == 'pct'
+    assert args.merge_threshold == 35

tests/test_merge_thresholds.py

@@ -0,0 +1,153 @@
+"""Test merge with thresholds
+"""
+
+from fediblockhole.blocklists import Blocklist, parse_blocklist
+from fediblockhole import merge_blocklists, apply_mergeplan
+
+from fediblockhole.const import SeverityLevel, DomainBlock
+
+datafile01 = "data-suspends-01.csv"
+datafile02 = "data-silences-01.csv"
+datafile03 = "data-noop-01.csv"
+
+import_fields = [
+    'domain',
+    'severity',
+    'public_comment',
+    'private_comment',
+    'reject_media',
+    'reject_reports',
+    'obfuscate'
+]
+
+def load_test_blocklist_data(datafiles):
+
+    blocklists = []
+
+    for df in datafiles:
+        with open(df) as fp:
+            data = fp.read()
+            bl = parse_blocklist(data, df, 'csv', import_fields)
+            blocklists.append(bl)
+    
+    return blocklists
+
+def test_mergeplan_count_2():
+    """Only merge a block if present in 2 or more lists
+    """
+
+    bl_1 = Blocklist('test01', {
+        'onemention.example.org': DomainBlock('onemention.example.org', 'suspend', '', '', True, True, True),
+        'twomention.example.org': DomainBlock('twomention.example.org', 'suspend', '', '', True, True, True),
+        'threemention.example.org': DomainBlock('threemention.example.org', 'suspend', '', '', True, True, True),
+        })
+
+    bl_2 = Blocklist('test2', {
+        'twomention.example.org': DomainBlock('twomention.example.org', 'suspend', '', '', True, True, True),
+        'threemention.example.org': DomainBlock('threemention.example.org', 'suspend', '', '', True, True, True),
+    })
+
+    bl_3 = Blocklist('test3', {
+        'threemention.example.org': DomainBlock('threemention.example.org', 'suspend', '', '', True, True, True),
+        'threemention.example.org': DomainBlock('threemention.example.org', 'suspend', '', '', True, True, True),
+    })
+
+    ml = merge_blocklists([bl_1, bl_2, bl_3], 'max', threshold=2)
+
+    assert 'onemention.example.org' not in ml
+    assert 'twomention.example.org' in ml
+    assert 'threemention.example.org' in ml
+
+def test_mergeplan_count_3():
+    """Only merge a block if present in 3 or more lists
+    """
+
+    bl_1 = Blocklist('test01', {
+        'onemention.example.org': DomainBlock('onemention.example.org', 'suspend', '', '', True, True, True),
+        'twomention.example.org': DomainBlock('twomention.example.org', 'suspend', '', '', True, True, True),
+        'threemention.example.org': DomainBlock('threemention.example.org', 'suspend', '', '', True, True, True),
+        })
+
+    bl_2 = Blocklist('test2', {
+        'twomention.example.org': DomainBlock('twomention.example.org', 'suspend', '', '', True, True, True),
+        'threemention.example.org': DomainBlock('threemention.example.org', 'suspend', '', '', True, True, True),
+    })
+
+    bl_3 = Blocklist('test3', {
+        'threemention.example.org': DomainBlock('threemention.example.org', 'suspend', '', '', True, True, True),
+        'threemention.example.org': DomainBlock('threemention.example.org', 'suspend', '', '', True, True, True),
+    })
+
+    ml = merge_blocklists([bl_1, bl_2, bl_3], 'max', threshold=3)
+
+    assert 'onemention.example.org' not in ml
+    assert 'twomention.example.org' not in ml
+    assert 'threemention.example.org' in ml
+
+def test_mergeplan_pct_30():
+    """Only merge a block if present in 2 or more lists
+    """
+
+    bl_1 = Blocklist('test01', {
+        'onemention.example.org': DomainBlock('onemention.example.org', 'suspend', '', '', True, True, True),
+        'twomention.example.org': DomainBlock('twomention.example.org', 'suspend', '', '', True, True, True),
+        'fourmention.example.org': DomainBlock('fourmention.example.org', 'suspend', '', '', True, True, True),
+
+        })
+
+    bl_2 = Blocklist('test2', {
+        'twomention.example.org': DomainBlock('twomention.example.org', 'suspend', '', '', True, True, True),
+        'threemention.example.org': DomainBlock('threemention.example.org', 'suspend', '', '', True, True, True),
+        'fourmention.example.org': DomainBlock('fourmention.example.org', 'suspend', '', '', True, True, True),
+    })
+
+    bl_3 = Blocklist('test3', {
+        'threemention.example.org': DomainBlock('threemention.example.org', 'suspend', '', '', True, True, True),
+        'fourmention.example.org': DomainBlock('fourmention.example.org', 'suspend', '', '', True, True, True),
+    })
+
+    bl_4 = Blocklist('test4', {
+        'threemention.example.org': DomainBlock('threemention.example.org', 'suspend', '', '', True, True, True),
+        'fourmention.example.org': DomainBlock('fourmention.example.org', 'suspend', '', '', True, True, True),
+    })
+
+    ml = merge_blocklists([bl_1, bl_2, bl_3, bl_4], 'max', threshold=30, threshold_type='pct')
+
+    assert 'onemention.example.org' not in ml
+    assert 'twomention.example.org' in ml
+    assert 'threemention.example.org' in ml
+    assert 'fourmention.example.org' in ml
+
+def test_mergeplan_pct_55():
+    """Only merge a block if present in 2 or more lists
+    """
+
+    bl_1 = Blocklist('test01', {
+        'onemention.example.org': DomainBlock('onemention.example.org', 'suspend', '', '', True, True, True),
+        'twomention.example.org': DomainBlock('twomention.example.org', 'suspend', '', '', True, True, True),
+        'fourmention.example.org': DomainBlock('fourmention.example.org', 'suspend', '', '', True, True, True),
+
+        })
+
+    bl_2 = Blocklist('test2', {
+        'twomention.example.org': DomainBlock('twomention.example.org', 'suspend', '', '', True, True, True),
+        'threemention.example.org': DomainBlock('threemention.example.org', 'suspend', '', '', True, True, True),
+        'fourmention.example.org': DomainBlock('fourmention.example.org', 'suspend', '', '', True, True, True),
+    })
+
+    bl_3 = Blocklist('test3', {
+        'threemention.example.org': DomainBlock('threemention.example.org', 'suspend', '', '', True, True, True),
+        'fourmention.example.org': DomainBlock('fourmention.example.org', 'suspend', '', '', True, True, True),
+    })
+
+    bl_4 = Blocklist('test4', {
+        'threemention.example.org': DomainBlock('threemention.example.org', 'suspend', '', '', True, True, True),
+        'fourmention.example.org': DomainBlock('fourmention.example.org', 'suspend', '', '', True, True, True),
+    })
+
+    ml = merge_blocklists([bl_1, bl_2, bl_3, bl_4], 'max', threshold=55, threshold_type='pct')
+
+    assert 'onemention.example.org' not in ml
+    assert 'twomention.example.org' not in ml
+    assert 'threemention.example.org' in ml
+    assert 'fourmention.example.org' in ml