my.github: some work in progress on generating consistent ids

sadly it seems that there are at several issues:

- gdpr has less detailed data so it's hard to generate a proper ID at times
- sometimes there is a small (1s?) discrepancy between created_at between same event in GDPR an API
- some API events can have duplicate payload, but different id, which violates uniqueness

my/github/common.py

@@ -1,6 +1,9 @@
 """
 Github events and their metadata: comments/issues/pull requests
 """
+from ..core import __NOT_HPI_MODULE__
+
+
 from datetime import datetime
 from typing import Optional, NamedTuple, Iterable, Set, Tuple
 
@@ -48,4 +51,12 @@ def parse_dt(s: str) -> datetime:
     return pytz.utc.localize(datetime.strptime(s, '%Y-%m-%dT%H:%M:%SZ'))
 
 
-from ..core import __NOT_HPI_MODULE__
+# experimental way of supportint event ids... not sure
+class EventIds:
+    @staticmethod
+    def repo_created(*, dts: str, name: str, ref_type: str, ref: Optional[str]) -> str:
+        return f'{dts}_repocreated_{name}_{ref_type}_{ref}'
+
+    @staticmethod
+    def pr(*, dts: str, action: str, url: str) -> str:
+        return f'{dts}_pr{action}_{url}'

my/github/gdpr.py

@@ -8,7 +8,7 @@ from typing import Iterable, Dict, Any
 from ..core.error import Res
 from ..core import get_files
 
-from .common import Event, parse_dt
+from .common import Event, parse_dt, EventIds
 
 # TODO later, use a separate user config? (github_gdpr)
 from my.config import github as user_config
@@ -87,11 +87,14 @@ def _parse_common(d: Dict) -> Dict:
 def _parse_repository(d: Dict) -> Event:
     pref = 'https://github.com/'
     url = d['url']
+    dts = d['created_at']
+    rt  = d['type']
     assert url.startswith(pref); name = url[len(pref):]
+    eid = EventIds.repo_created(dts=dts, name=name, ref_type=rt, ref=None)
     return Event( # type: ignore[misc]
         **_parse_common(d),
         summary='created ' + name,
-        eid='created_' + name, # TODO ??
+        eid=eid,
     )
 
 
@@ -119,6 +122,7 @@ def _parse_issue(d: Dict) -> Event:
 
 
 def _parse_pull_request(d: Dict) -> Event:
+    dts = d['created_at']
     url = d['url']
     title = d['title']
     is_bot = "[bot]" in d["user"]
@@ -127,7 +131,7 @@ def _parse_pull_request(d: Dict) -> Event:
         # TODO distinguish incoming/outgoing?
         # TODO action? opened/closed??
         summary=f'opened PR {title}',
-        eid='pull_request_' + url,
+        eid=EventIds.pr(dts=dts, action='opened', url=url),
         is_bot=is_bot,
     )
 
@@ -146,6 +150,7 @@ def _parse_project(d: Dict) -> Event:
         is_bot=is_bot,
     )
 
+
 def _parse_release(d: Dict) -> Event:
     tag = d['tag_name']
     return Event( # type: ignore[misc]

my/github/ghexport.py

@@ -46,7 +46,7 @@ from typing import Tuple, Dict, Sequence, Optional
 from my.core import get_files, Path, LazyLogger
 from my.core.common import mcachew
 
-from .common import Event, parse_dt, Results
+from .common import Event, parse_dt, Results, EventIds
 
 
 logger = LazyLogger(__name__)
@@ -61,9 +61,16 @@ def _dal() -> dal.DAL:
     return dal.DAL(sources)
 
 
-# todo cachew: hmm. not good, need to be lazier?...
 @mcachew(depends_on=lambda: inputs())
 def events() -> Results:
+    from my.core.common import ensure_unique
+    key = lambda e: object() if isinstance(e, Exception) else e.eid
+    # crap. sometimes API events can be repeated with exactly the same payload and different id
+    # yield from ensure_unique(_events(), key=key)
+    yield from _events()
+
+
+def _events() -> Results:
     dal = _dal()
     for d in dal.events():
         if isinstance(d, Exception):
@@ -93,6 +100,7 @@ EventId = str
 Body = str
 def _get_summary(e) -> Tuple[str, Optional[Link], Optional[EventId], Optional[Body]]:
     # TODO would be nice to give access to raw event within timeline
+    dts = e['created_at']
     eid = e['id']
     tp = e['type']
     pl = e['payload']
@@ -119,10 +127,13 @@ def _get_summary(e) -> Tuple[str, Optional[Link], Optional[EventId], Optional[Bo
         what = mapping[tp]
         rt  = pl['ref_type']
         ref = pl['ref']
-        # TODO link to branch? only contains weird API link though
+        if what == 'created':
-        # TODO hmm. include timestamp instead?
+            # FIXME should handle delection?...
+            eid = EventIds.repo_created(dts=dts, name=rname, ref_type=rt, ref=ref)
+        mref = '' if ref is None else ' ' + ref
+        # todo link to branch? only contains weird API link though
         # TODO combine automatically instead
-        return f"{rname}: {what} {rt} {ref}", None, f'{rname}_{what}_{rt}_{ref}_{eid}', None
+        return f"{rname}: {what} {rt}{mref}", None, eid, None
     elif tp == 'PullRequestEvent':
         pr = pl['pull_request']
         title = pr['title']
@@ -130,7 +141,8 @@ def _get_summary(e) -> Tuple[str, Optional[Link], Optional[EventId], Optional[Bo
         link  = pr['html_url']
         body  = pr['body']
         action = pl['action']
-        return f"{rname}: {action} PR: {title}", link, f'{rname}_{action}_pr_{link}', body
+        eid = EventIds.pr(dts=dts, action=action, url=link)
+        return f"{rname}: {action} PR: {title}", link, eid, body
     elif tp == 'PullRequestReviewEvent':
         pr = pl['pull_request']
         title = pr['title']