use updated doublethink library populate_defaults() to avoid problem where under certain circumstances field values from the database would be overwritten by defaults

brozzler/job.py

@@ -71,8 +71,8 @@ def new_job(frontier, job_conf):
     '''Returns new Job.'''
     validate_conf(job_conf)
     job = Job(frontier.rr, {
-                "conf": job_conf,
+                "conf": job_conf, "status": "ACTIVE",
-                "status": "ACTIVE", "started": doublethink.utcnow()})
+                "started": doublethink.utcnow()})
     if "id" in job_conf:
         job.id = job_conf["id"]
     job.save()
@@ -80,22 +80,9 @@ def new_job(frontier, job_conf):
     sites = []
     for seed_conf in job_conf["seeds"]:
         merged_conf = merge(seed_conf, job_conf)
-        site = brozzler.Site(frontier.rr, {
+        merged_conf["job_id"] = job.id
-            "job_id": job.id,
+        merged_conf["seed"] = merged_conf.pop("url")
-            "seed": merged_conf["url"],
+        site = brozzler.Site(frontier.rr, merged_conf)
-            "scope": merged_conf.get("scope"),
-            "time_limit": merged_conf.get("time_limit"),
-            "proxy": merged_conf.get("proxy"),
-            "ignore_robots": merged_conf.get("ignore_robots"),
-            "enable_warcprox_features": merged_conf.get(
-                "enable_warcprox_features"),
-            "warcprox_meta": merged_conf.get("warcprox_meta"),
-            "metadata": merged_conf.get("metadata"),
-            "remember_outlinks": merged_conf.get("remember_outlinks"),
-            "user_agent": merged_conf.get("user_agent"),
-            "behavior_parameters": merged_conf.get("behavior_parameters"),
-            "username": merged_conf.get("username"),
-            "password": merged_conf.get("password")})
         sites.append(site)
 
     for site in sites:
@@ -127,9 +114,9 @@ class Job(doublethink.Document):
     logger = logging.getLogger(__module__ + "." + __qualname__)
     table = "jobs"
 
-    def __init__(self, rr, d={}):
+    def populate_defaults(self):
-        doublethink.Document.__init__(self, rr, d)
+        if not "status" in self:
-        self.status = self.get("status", "ACTIVE")
+            self.status = "ACTIVE"
         if not "starts_and_stops" in self:
             if self.get("started"):   # backward compatibility
                 self.starts_and_stops = [{

brozzler/site.py

@@ -92,28 +92,32 @@ class Site(doublethink.Document):
     logger = logging.getLogger(__module__ + "." + __qualname__)
     table = 'sites'
 
-    def __init__(self, rr, d={}):
+    def populate_defaults(self):
-        doublethink.Document.__init__(self, rr, d)
+        if not "status" in self:
-        if not self.get('status'):
+            self.status = "ACTIVE"
-            self.status = 'ACTIVE'
+        if not "enable_warcprox_features" in self:
-        self.enable_warcprox_features = bool(self.get('enable_warcprox_features'))
+            self.enable_warcprox_features = False
-        self.claimed = bool(self.get('claimed'))
+        if not "claimed" in self:
-        self.last_disclaimed = self.get('last_disclaimed', _EPOCH_UTC)
+            self.claimed = False
-        self.last_claimed = self.get('last_claimed', _EPOCH_UTC)
+        if not "last_disclaimed" in self:
-        if not self.get('starts_and_stops'):
+            self.last_disclaimed = _EPOCH_UTC
-            if self.get('start_time'):   # backward compatibility
+        if not "last_claimed" in self:
+            self.last_claimed = _EPOCH_UTC
+        if not "scope" in self:
+            self.scope = {}
+        if not "surt" in self.scope:
+            self.scope["surt"] = Url(self.seed).surt
+
+        if not "starts_and_stops" in self:
+            if self.get("start_time"):   # backward compatibility
                 self.starts_and_stops = [{
-                    'start':self.get('start_time'),'stop':None}]
+                    "start":self.get("start_time"),"stop":None}]
-                if self.get('status') != 'ACTIVE':
+                if self.get("status") != "ACTIVE":
-                    self.starts_and_stops[0]['stop'] = self.last_disclaimed
+                    self.starts_and_stops[0]["stop"] = self.last_disclaimed
-                del self['start_time']
+                del self["start_time"]
             else:
                 self.starts_and_stops = [
-                        {'start':doublethink.utcnow(),'stop':None}]
+                        {"start":doublethink.utcnow(),"stop":None}]
-        if not self.scope:
-            self.scope = {}
-        if not 'surt' in self.scope:
-            self.scope['surt'] = Url(self.seed).surt
 
     def __str__(self):
         return 'Site({"id":"%s","seed":"%s",...})' % (self.id, self.seed)
@@ -286,19 +290,22 @@ class Site(doublethink.Document):
 
 class Page(doublethink.Document):
     logger = logging.getLogger(__module__ + "." + __qualname__)
-    table = 'pages'
+    table = "pages"
 
-    def __init__(self, rr, d={}):
+    def populate_defaults(self):
-        doublethink.Document.__init__(self, rr, d)
+        if not "hops_from_seed" in self:
-        self.hops_from_seed = self.get('hops_from_seed', 0)
+            self.hops_from_seed = 0
-        self.brozzle_count = self.get('brozzle_count', 0)
+        if not "brozzle_count" in self:
-        self.claimed = self.get('claimed', False)
+            self.brozzle_count = 0
-        self.hops_off_surt = self.get('hops_off_surt', 0)
+        if not "claimed" in self:
-        self.needs_robots_check = self.get('needs_robots_check', False)
+            self.claimed = False
-        self._canon_hurl = None
+        if not "hops_off_surt" in self:
-
+            self.hops_off_surt = 0
-        self.priority = self.get('priority', self._calc_priority())
+        if not "needs_robots_check" in self:
-        if self.get('id') is None:
+            self.needs_robots_check = False
+        if not "priority" in self:
+            self.priority = self._calc_priority()
+        if not "id" in self:
             digest_this = "site_id:%s,url:%s" % (self.site_id, self.url)
             self.id = hashlib.sha1(digest_this.encode("utf-8")).hexdigest()
 

setup.py

@@ -32,7 +32,7 @@ def find_package_data(package):
 
 setuptools.setup(
         name='brozzler',
-        version='1.1b9.dev199',
+        version='1.1b9.dev200',
         description='Distributed web crawling with browsers',
         url='https://github.com/internetarchive/brozzler',
         author='Noah Levitt',
@@ -69,7 +69,7 @@ setuptools.setup(
             'websocket-client!=0.39.0',
             'pillow==3.3.0',
             'surt>=0.3.0',
-            'doublethink>=0.2.0.dev66',
+            'doublethink>=0.2.0.dev70',
             'rethinkdb>=2.3,<2.4',
             'cerberus==1.0.1',
             'jinja2',

tests/test_frontier.py

@@ -172,3 +172,78 @@ def test_time_limit():
     assert site.starts_and_stops[1]['start']
     assert site.starts_and_stops[1]['stop']
     assert site.starts_and_stops[1]['stop'] > site.starts_and_stops[0]['start']
+
+def test_field_defaults():
+    rr = doublethink.Rethinker('localhost', db='ignoreme')
+
+    # page
+    brozzler.Page.table_ensure(rr)
+    page = brozzler.Page(rr, {'hops_from_seed': 3})
+    assert page.hops_from_seed == 3
+    assert page.id is None
+    assert page.brozzle_count is None
+    page.save()
+    assert page.hops_from_seed == 3
+    assert page.id
+    assert page.brozzle_count == 0
+
+    qage = brozzler.Page.load(rr, page.id)
+    assert qage.hops_from_seed == 3
+    assert qage.id == page.id
+    assert qage.brozzle_count == 0
+    qage.save()
+    assert qage.hops_from_seed == 3
+    assert qage.id == page.id
+    assert qage.brozzle_count == 0
+    qage.refresh()
+    assert qage.hops_from_seed == 3
+    assert qage.id == page.id
+    assert qage.brozzle_count == 0
+
+    # site
+    brozzler.Site.table_ensure(rr)
+    site = brozzler.Site(rr, {'enable_warcprox_features': True})
+    assert site.enable_warcprox_features is True
+    assert site.id is None
+    assert site.scope is None
+    site.save()
+    assert site.id
+    assert site.scope
+
+    tite = brozzler.Site.load(rr, site.id)
+    assert tite.enable_warcprox_features is True
+    assert tite.id == site.id
+    assert tite.scope == site.scope
+    tite.save()
+    assert tite.enable_warcprox_features is True
+    assert tite.id == site.id
+    assert tite.scope == site.scope
+    tite.refresh()
+    assert tite.enable_warcprox_features is True
+    assert tite.id == site.id
+    assert tite.scope == site.scope
+
+    # job
+    brozzler.Job.table_ensure(rr)
+    job = brozzler.Job(rr, {'status': 'WHUUUT'})
+    assert job.status == 'WHUUUT'
+    assert job.id is None
+    assert job.starts_and_stops is None
+    job.save()
+    assert job.status == 'WHUUUT'
+    assert job.id
+    assert job.starts_and_stops
+
+    kob = brozzler.Job.load(rr, job.id)
+    assert kob.status == 'WHUUUT'
+    assert kob.id
+    assert kob.starts_and_stops
+    kob.save()
+    assert kob.status == 'WHUUUT'
+    assert kob.id
+    assert kob.starts_and_stops
+    kob.refresh()
+    assert kob.status == 'WHUUUT'
+    assert kob.id
+    assert kob.starts_and_stops
+