feat: archive git repository (experimental)

See doc/git-archive.md for general Git archive specifications
See doc/repos/metadata-repo.md for info and direction related to the new Git metadata archive

aurweb/archives/__init__.py

@@ -0,0 +1 @@
+# aurweb.archives

aurweb/archives/spec/__init__.py

@@ -0,0 +1 @@
+# aurweb.archives.spec

aurweb/archives/spec/base.py

@@ -0,0 +1,77 @@
+from pathlib import Path
+from typing import Any, Dict, Iterable, List, Set
+
+
+class GitInfo:
+    """Information about a Git repository."""
+
+    """ Path to Git repository. """
+    path: str
+
+    """ Local Git repository configuration. """
+    config: Dict[str, Any]
+
+    def __init__(self, path: str, config: Dict[str, Any] = dict()) -> "GitInfo":
+        self.path = Path(path)
+        self.config = config
+
+
+class SpecOutput:
+    """Class used for git_archive.py output details."""
+
+    """ Filename relative to the Git repository root. """
+    filename: Path
+
+    """ Git repository information. """
+    git_info: GitInfo
+
+    """ Bytes bound for `SpecOutput.filename`. """
+    data: bytes
+
+    def __init__(self, filename: str, git_info: GitInfo, data: bytes) -> "SpecOutput":
+        self.filename = filename
+        self.git_info = git_info
+        self.data = data
+
+
+class SpecBase:
+    """
+    Base for Spec classes defined in git_archve.py --spec modules.
+
+    All supported --spec modules must contain the following classes:
+    - Spec(SpecBase)
+    """
+
+    """ A list of SpecOutputs, each of which contain output file data. """
+    outputs: List[SpecOutput] = list()
+
+    """ A set of repositories to commit changes to. """
+    repos: Set[str] = set()
+
+    def generate(self) -> Iterable[SpecOutput]:
+        """
+        "Pure virtual" output generator.
+
+        `SpecBase.outputs` and `SpecBase.repos` should be populated within an
+        overridden version of this function in SpecBase derivatives.
+        """
+        raise NotImplementedError()
+
+    def add_output(self, filename: str, git_info: GitInfo, data: bytes) -> None:
+        """
+        Add a SpecOutput instance to the set of outputs.
+
+        :param filename: Filename relative to the git repository root
+        :param git_info: GitInfo instance
+        :param data: Binary data bound for `filename`
+        """
+        if git_info.path not in self.repos:
+            self.repos.add(git_info.path)
+
+        self.outputs.append(
+            SpecOutput(
+                filename,
+                git_info,
+                data,
+            )
+        )

aurweb/archives/spec/metadata.py

@@ -0,0 +1,85 @@
+from typing import Iterable
+
+import orjson
+
+from aurweb import config, db
+from aurweb.models import Package, PackageBase, User
+from aurweb.rpc import RPC
+
+from .base import GitInfo, SpecBase, SpecOutput
+
+ORJSON_OPTS = orjson.OPT_SORT_KEYS | orjson.OPT_INDENT_2
+
+
+class Spec(SpecBase):
+    def __init__(self) -> "Spec":
+        self.metadata_repo = GitInfo(
+            config.get("git-archive", "metadata-repo"),
+        )
+
+    def generate(self) -> Iterable[SpecOutput]:
+        # Base query used by the RPC.
+        base_query = (
+            db.query(Package)
+            .join(PackageBase)
+            .join(User, PackageBase.MaintainerUID == User.ID)
+        )
+
+        # Create an instance of RPC, use it to get entities from
+        # our query and perform a metadata subquery for all packages.
+        rpc = RPC(version=5, type="info")
+        print("performing package database query")
+        packages = rpc.entities(base_query).all()
+        print("performing package database subqueries")
+        rpc.subquery({pkg.ID for pkg in packages})
+
+        pkgbases, pkgnames = dict(), dict()
+        for package in packages:
+            # Produce RPC type=info data for `package`
+            data = rpc.get_info_json_data(package)
+
+            pkgbase_name = data.get("PackageBase")
+            pkgbase_data = {
+                "ID": data.pop("PackageBaseID"),
+                "URLPath": data.pop("URLPath"),
+                "FirstSubmitted": data.pop("FirstSubmitted"),
+                "LastModified": data.pop("LastModified"),
+                "OutOfDate": data.pop("OutOfDate"),
+                "Maintainer": data.pop("Maintainer"),
+                "Keywords": data.pop("Keywords"),
+                "NumVotes": data.pop("NumVotes"),
+                "Popularity": data.pop("Popularity"),
+                "PopularityUpdated": package.PopularityUpdated.timestamp(),
+            }
+
+            # Store the data in `pkgbases` dict. We do this so we only
+            # end up processing a single `pkgbase` if repeated after
+            # this loop
+            pkgbases[pkgbase_name] = pkgbase_data
+
+            # Remove Popularity and NumVotes from package data.
+            # These fields change quite often which causes git data
+            # modification to explode.
+            # data.pop("NumVotes")
+            # data.pop("Popularity")
+
+            # Remove the ID key from package json.
+            data.pop("ID")
+
+            # Add the `package`.Name to the pkgnames set
+            name = data.get("Name")
+            pkgnames[name] = data
+
+        # Add metadata outputs
+        self.add_output(
+            "pkgname.json",
+            self.metadata_repo,
+            orjson.dumps(pkgnames, option=ORJSON_OPTS),
+        )
+        self.add_output(
+            "pkgbase.json",
+            self.metadata_repo,
+            orjson.dumps(pkgbases, option=ORJSON_OPTS),
+        )
+
+        return self.outputs

aurweb/archives/spec/pkgbases.py

@@ -0,0 +1,32 @@
+from typing import Iterable
+
+import orjson
+
+from aurweb import config, db
+from aurweb.models import PackageBase
+
+from .base import GitInfo, SpecBase, SpecOutput
+
+ORJSON_OPTS = orjson.OPT_SORT_KEYS | orjson.OPT_INDENT_2
+
+
+class Spec(SpecBase):
+    def __init__(self) -> "Spec":
+        self.pkgbases_repo = GitInfo(config.get("git-archive", "pkgbases-repo"))
+
+    def generate(self) -> Iterable[SpecOutput]:
+        filt = PackageBase.PackagerUID.isnot(None)
+        query = (
+            db.query(PackageBase.Name)
+            .filter(filt)
+            .order_by(PackageBase.Name.asc())
+            .all()
+        )
+        pkgbases = [pkgbase.Name for pkgbase in query]
+
+        self.add_output(
+            "pkgbase.json",
+            self.pkgbases_repo,
+            orjson.dumps(pkgbases, option=ORJSON_OPTS),
+        )
+        return self.outputs

aurweb/archives/spec/pkgnames.py

@@ -0,0 +1,33 @@
+from typing import Iterable
+
+import orjson
+
+from aurweb import config, db
+from aurweb.models import Package, PackageBase
+
+from .base import GitInfo, SpecBase, SpecOutput
+
+ORJSON_OPTS = orjson.OPT_SORT_KEYS | orjson.OPT_INDENT_2
+
+
+class Spec(SpecBase):
+    def __init__(self) -> "Spec":
+        self.pkgnames_repo = GitInfo(config.get("git-archive", "pkgnames-repo"))
+
+    def generate(self) -> Iterable[SpecOutput]:
+        filt = PackageBase.PackagerUID.isnot(None)
+        query = (
+            db.query(Package.Name)
+            .join(PackageBase, PackageBase.ID == Package.PackageBaseID)
+            .filter(filt)
+            .order_by(Package.Name.asc())
+            .all()
+        )
+        pkgnames = [pkg.Name for pkg in query]
+
+        self.add_output(
+            "pkgname.json",
+            self.pkgnames_repo,
+            orjson.dumps(pkgnames, option=ORJSON_OPTS),
+        )
+        return self.outputs

aurweb/archives/spec/users.py

@@ -0,0 +1,26 @@
+from typing import Iterable
+
+import orjson
+
+from aurweb import config, db
+from aurweb.models import User
+
+from .base import GitInfo, SpecBase, SpecOutput
+
+ORJSON_OPTS = orjson.OPT_SORT_KEYS | orjson.OPT_INDENT_2
+
+
+class Spec(SpecBase):
+    def __init__(self) -> "Spec":
+        self.users_repo = GitInfo(config.get("git-archive", "users-repo"))
+
+    def generate(self) -> Iterable[SpecOutput]:
+        query = db.query(User.Username).order_by(User.Username.asc()).all()
+        users = [user.Username for user in query]
+
+        self.add_output(
+            "users.json",
+            self.users_repo,
+            orjson.dumps(users, option=ORJSON_OPTS),
+        )
+        return self.outputs

aurweb/models/package_base.py

@@ -64,3 +64,13 @@ class PackageBase(Base):
         if key in PackageBase.TO_FLOAT and not isinstance(attr, float):
             return float(attr)
         return attr
+
+
+def popularity_decay(pkgbase: PackageBase, utcnow: int):
+    """Return the delta between now and the last time popularity was updated, in days"""
+    return int((utcnow - pkgbase.PopularityUpdated.timestamp()) / 86400)
+
+
+def popularity(pkgbase: PackageBase, utcnow: int):
+    """Return up-to-date popularity"""
+    return float(pkgbase.Popularity) * (0.98 ** popularity_decay(pkgbase, utcnow))

aurweb/pkgbase/util.py

@@ -3,8 +3,9 @@ from typing import Any
 from fastapi import Request
 from sqlalchemy import and_
 
-from aurweb import config, db, defaults, l10n, util
+from aurweb import config, db, defaults, l10n, time, util
 from aurweb.models import PackageBase, User
+from aurweb.models.package_base import popularity
 from aurweb.models.package_comaintainer import PackageComaintainer
 from aurweb.models.package_comment import PackageComment
 from aurweb.models.package_request import PENDING_ID, PackageRequest
@@ -81,6 +82,8 @@ def make_context(
         and_(PackageRequest.Status == PENDING_ID, PackageRequest.ClosedTS.is_(None))
     ).count()
 
+    context["popularity"] = popularity(pkgbase, time.utcnow())
+
     return context
 
 

aurweb/rpc.py

@@ -6,9 +6,10 @@ from fastapi.responses import HTMLResponse
 from sqlalchemy import and_, literal, orm
 
 import aurweb.config as config
-from aurweb import db, defaults, models
+from aurweb import db, defaults, models, time
 from aurweb.exceptions import RPCError
 from aurweb.filters import number_format
+from aurweb.models.package_base import popularity
 from aurweb.packages.search import RPCSearch
 
 TYPE_MAPPING = {
@@ -120,16 +121,15 @@ class RPC:
         if not args:
             raise RPCError("No request type/data specified.")
 
-    def _get_json_data(self, package: models.Package) -> dict[str, Any]:
+    def get_json_data(self, package: models.Package) -> dict[str, Any]:
         """Produce dictionary data of one Package that can be JSON-serialized.
 
         :param package: Package instance
         :returns: JSON-serializable dictionary
         """
 
-        # Produce RPC API compatible Popularity: If zero, it's an integer
+        # Normalize Popularity for RPC output to 6 decimal precision
-        # 0, otherwise, it's formatted to the 6th decimal place.
+        pop = popularity(package, time.utcnow())
-        pop = package.Popularity
         pop = 0 if not pop else float(number_format(pop, 6))
 
         snapshot_uri = config.get("options", "snapshot_uri")
@@ -151,8 +151,8 @@ class RPC:
             "LastModified": package.ModifiedTS,
         }
 
-    def _get_info_json_data(self, package: models.Package) -> dict[str, Any]:
+    def get_info_json_data(self, package: models.Package) -> dict[str, Any]:
-        data = self._get_json_data(package)
+        data = self.get_json_data(package)
 
         # All info results have _at least_ an empty list of
         # License and Keywords.
@@ -176,7 +176,7 @@ class RPC:
         """
         return [data_generator(pkg) for pkg in packages]
 
-    def _entities(self, query: orm.Query) -> orm.Query:
+    def entities(self, query: orm.Query) -> orm.Query:
         """Select specific RPC columns on `query`."""
         return query.with_entities(
             models.Package.ID,
@@ -188,38 +188,14 @@ class RPC:
             models.PackageBase.Name.label("PackageBaseName"),
             models.PackageBase.NumVotes,
             models.PackageBase.Popularity,
+            models.PackageBase.PopularityUpdated,
             models.PackageBase.OutOfDateTS,
             models.PackageBase.SubmittedTS,
             models.PackageBase.ModifiedTS,
             models.User.Username.label("Maintainer"),
         ).group_by(models.Package.ID)
 
-    def _handle_multiinfo_type(
+    def subquery(self, ids: set[int]):
-        self, args: list[str] = [], **kwargs
-    ) -> list[dict[str, Any]]:
-        self._enforce_args(args)
-        args = set(args)
-
-        packages = (
-            db.query(models.Package)
-            .join(models.PackageBase)
-            .join(
-                models.User,
-                models.User.ID == models.PackageBase.MaintainerUID,
-                isouter=True,
-            )
-            .filter(models.Package.Name.in_(args))
-        )
-
-        max_results = config.getint("options", "max_rpc_results")
-        packages = self._entities(packages).limit(max_results + 1)
-
-        if packages.count() > max_results:
-            raise RPCError("Too many package results.")
-
-        ids = {pkg.ID for pkg in packages}
-
-        # Aliases for 80-width.
         Package = models.Package
         PackageKeyword = models.PackageKeyword
 
@@ -311,7 +287,33 @@ class RPC:
 
             self.extra_info[record.ID][type_].append(name)
 
-        return self._assemble_json_data(packages, self._get_info_json_data)
+    def _handle_multiinfo_type(
+        self, args: list[str] = [], **kwargs
+    ) -> list[dict[str, Any]]:
+        self._enforce_args(args)
+        args = set(args)
+
+        packages = (
+            db.query(models.Package)
+            .join(models.PackageBase)
+            .join(
+                models.User,
+                models.User.ID == models.PackageBase.MaintainerUID,
+                isouter=True,
+            )
+            .filter(models.Package.Name.in_(args))
+        )
+
+        max_results = config.getint("options", "max_rpc_results")
+        packages = self.entities(packages).limit(max_results + 1)
+
+        if packages.count() > max_results:
+            raise RPCError("Too many package results.")
+
+        ids = {pkg.ID for pkg in packages}
+        self.subquery(ids)
+
+        return self._assemble_json_data(packages, self.get_info_json_data)
 
     def _handle_search_type(
         self, by: str = defaults.RPC_SEARCH_BY, args: list[str] = []
@@ -330,12 +332,12 @@ class RPC:
         search.search_by(by, arg)
 
         max_results = config.getint("options", "max_rpc_results")
-        results = self._entities(search.results()).limit(max_results + 1).all()
+        results = self.entities(search.results()).limit(max_results + 1).all()
 
         if len(results) > max_results:
             raise RPCError("Too many package results.")
 
-        return self._assemble_json_data(results, self._get_json_data)
+        return self._assemble_json_data(results, self.get_json_data)
 
     def _handle_msearch_type(
         self, args: list[str] = [], **kwargs

aurweb/schema.py

@@ -155,6 +155,12 @@ PackageBases = Table(
         nullable=False,
         server_default=text("0"),
     ),
+    Column(
+        "PopularityUpdated",
+        TIMESTAMP,
+        nullable=False,
+        server_default=text("'1970-01-01 00:00:01.000000'"),
+    ),
     Column("OutOfDateTS", BIGINT(unsigned=True)),
     Column("FlaggerComment", Text, nullable=False),
     Column("SubmittedTS", BIGINT(unsigned=True), nullable=False),

aurweb/scripts/git_archive.py

@@ -0,0 +1,125 @@
+import argparse
+import importlib
+import os
+import sys
+import traceback
+from datetime import datetime
+
+import orjson
+import pygit2
+
+from aurweb import config
+
+# Constants
+REF = "refs/heads/master"
+ORJSON_OPTS = orjson.OPT_SORT_KEYS | orjson.OPT_INDENT_2
+
+
+def init_repository(git_info) -> None:
+    pygit2.init_repository(git_info.path)
+    repo = pygit2.Repository(git_info.path)
+    for k, v in git_info.config.items():
+        repo.config[k] = v
+
+
+def parse_args():
+    parser = argparse.ArgumentParser()
+    parser.add_argument(
+        "--spec",
+        type=str,
+        required=True,
+        help="name of spec module in the aurweb.archives.spec package",
+    )
+    return parser.parse_args()
+
+
+def update_repository(repo: pygit2.Repository):
+    # Use git status to determine file changes
+    has_changes = False
+    changes = repo.status()
+    for filepath, flags in changes.items():
+        if flags != pygit2.GIT_STATUS_CURRENT:
+            has_changes = True
+            break
+
+    if has_changes:
+        print("diff detected, committing")
+        # Add everything in the tree.
+        print("adding files to git tree")
+
+        # Add the tree to staging
+        repo.index.read()
+        repo.index.add_all()
+        repo.index.write()
+        tree = repo.index.write_tree()
+
+        # Determine base commit; if repo.head.target raises GitError,
+        # we have no current commits
+        try:
+            base = [repo.head.target]
+        except pygit2.GitError:
+            base = []
+
+        utcnow = datetime.utcnow()
+        author = pygit2.Signature(
+            config.get("git-archive", "author"),
+            config.get("git-archive", "author-email"),
+            int(utcnow.timestamp()),
+            0,
+        )
+
+        # Commit the changes
+        timestamp = utcnow.strftime("%Y-%m-%d %H:%M:%S")
+        title = f"update - {timestamp}"
+        repo.create_commit(REF, author, author, title, tree, base)
+
+        print("committed changes")
+    else:
+        print("no diff detected")
+
+
+def main() -> int:
+    args = parse_args()
+
+    print(f"loading '{args.spec}' spec")
+    spec_package = "aurweb.archives.spec"
+    module_path = f"{spec_package}.{args.spec}"
+    spec_module = importlib.import_module(module_path)
+    print(f"loaded '{args.spec}'")
+
+    # Track repositories that the spec modifies. After we run
+    # through specs, we want to make a single commit for all
+    # repositories that contain changes.
+    repos = dict()
+
+    print(f"running '{args.spec}' spec...")
+    spec = spec_module.Spec()
+    for output in spec.generate():
+        if not os.path.exists(output.git_info.path / ".git"):
+            init_repository(output.git_info)
+
+        path = output.git_info.path / output.filename
+        with open(path, "wb") as f:
+            f.write(output.data)
+
+        if output.git_info.path not in repos:
+            repos[output.git_info.path] = pygit2.Repository(output.git_info.path)
+
+    print(f"done running '{args.spec}' spec")
+
+    print("processing repositories")
+    for path in spec.repos:
+        print(f"processing repository: {path}")
+        update_repository(pygit2.Repository(path))
+
+    return 0
+
+
+if __name__ == "__main__":
+    try:
+        sys.exit(main())
+    except KeyboardInterrupt:
+        sys.exit(0)
+    except Exception:
+        traceback.print_exc()
+        sys.exit(1)

aurweb/scripts/mkpkglists.py

@@ -188,6 +188,7 @@ def _main():
     USERS = aurweb.config.get("mkpkglists", "userfile")
 
     bench = Benchmark()
+    logger.warning(f"{sys.argv[0]} is deprecated and will be soon be removed")
     logger.info("Started re-creating archives, wait a while...")
 
     query = (

aurweb/scripts/popupdate.py

@@ -1,9 +1,10 @@
 #!/usr/bin/env python3
+from datetime import datetime
 
 from sqlalchemy import and_, func
 from sqlalchemy.sql.functions import coalesce, sum as _sum
 
-from aurweb import db, time
+from aurweb import config, db, time
 from aurweb.models import PackageBase, PackageVote
 
 
@@ -46,13 +47,24 @@ def run_variable(pkgbases: list[PackageBase] = []) -> None:
 
         ids = set()
         if pkgbases:
+            # If `pkgbases` were given, we should forcefully update the given
+            # package base records' popularities.
             ids = {pkgbase.ID for pkgbase in pkgbases}
             query = query.filter(PackageBase.ID.in_(ids))
+        else:
+            # Otherwise, we should only update popularities which have exceeded
+            # the popularity interval length.
+            interval = config.getint("git-archive", "popularity-interval")
+            query = query.filter(
+                PackageBase.PopularityUpdated
+                <= datetime.fromtimestamp((now - interval))
+            )
 
         query.update(
             {
                 "NumVotes": votes_subq.scalar_subquery(),
                 "Popularity": pop_subq.scalar_subquery(),
+                "PopularityUpdated": datetime.fromtimestamp(now),
             }
         )
 

aurweb/testing/git.py

@@ -1,6 +1,4 @@
 import os
-import shlex
-from subprocess import PIPE, Popen
 from typing import Tuple
 
 import py
@@ -8,6 +6,7 @@ import py
 from aurweb.models import Package
 from aurweb.templates import base_template
 from aurweb.testing.filelock import FileLock
+from aurweb.util import shell_exec
 
 
 class GitRepository:
@@ -24,10 +23,7 @@ class GitRepository:
         self.file_lock.lock(on_create=self._setup)
 
     def _exec(self, cmdline: str, cwd: str) -> Tuple[int, str, str]:
-        args = shlex.split(cmdline)
+        return shell_exec(cmdline, cwd)
-        proc = Popen(args, cwd=cwd, stdout=PIPE, stderr=PIPE)
-        out, err = proc.communicate()
-        return (proc.returncode, out.decode().strip(), err.decode().strip())
 
     def _exec_repository(self, cmdline: str) -> Tuple[int, str, str]:
         return self._exec(cmdline, cwd=str(self.file_lock.path))

aurweb/util.py

@@ -1,6 +1,7 @@
 import math
 import re
 import secrets
+import shlex
 import string
 from datetime import datetime
 from http import HTTPStatus
@@ -192,3 +193,10 @@ def parse_ssh_key(string: str) -> Tuple[str, str]:
 def parse_ssh_keys(string: str) -> list[Tuple[str, str]]:
     """Parse a list of SSH public keys."""
     return [parse_ssh_key(e) for e in string.splitlines()]
+
+
+def shell_exec(cmdline: str, cwd: str) -> Tuple[int, str, str]:
+    args = shlex.split(cmdline)
+    proc = Popen(args, cwd=cwd, stdout=PIPE, stderr=PIPE)
+    out, err = proc.communicate()
+    return (proc.returncode, out.decode().strip(), err.decode().strip())

conf/config.defaults

@@ -131,6 +131,18 @@ packagesmetaextfile = /srv/http/aurweb/web/html/packages-meta-ext-v1.json.gz
 pkgbasefile = /srv/http/aurweb/web/html/pkgbase.gz
 userfile = /srv/http/aurweb/web/html/users.gz
 
+[git-archive]
+author = git_archive.py
+author-email = no-reply@archlinux.org
+
+; One week worth of seconds (86400 * 7)
+popularity-interval = 604800
+
+metadata-repo = /srv/http/aurweb/metadata.git
+users-repo = /srv/http/aurweb/users.git
+pkgbases-repo = /srv/http/aurweb/pkgbases.git
+pkgnames-repo = /srv/http/aurweb/pkgnames.git
+
 [devel]
 ; commit_url is a format string used to produce a link to a commit hash.
 commit_url = https://gitlab.archlinux.org/archlinux/aurweb/-/commits/%s

conf/config.dev

@@ -76,5 +76,11 @@ packagesmetaextfile = /var/lib/aurweb/archives/packages-meta-ext-v1.json.gz
 pkgbasefile = /var/lib/aurweb/archives/pkgbase.gz
 userfile = /var/lib/aurweb/archives/users.gz
 
+[git-archive]
+metadata-repo = metadata.git
+users-repo = users.git
+pkgbases-repo = pkgbases.git
+pkgnames-repo = pkgnames.git
+
 [aurblup]
 db-path = YOUR_AUR_ROOT/aurblup/

doc/git-archive.md

@@ -0,0 +1,75 @@
+# aurweb Git Archive Specification
+
+<span style="color: red">
+    WARNING: This aurweb Git Archive implementation is
+    experimental and may be changed.
+</span>
+
+## Overview
+
+This git archive specification refers to the archive git repositories
+created by [aurweb/scripts/git_archive.py](aurweb/scripts/git_archive.py)
+using [spec modules](#spec-modules).
+
+## Configuration
+
+- `[git-archive]`
+    - `author`
+        - Git commit author
+    - `author-email`
+        - Git commit author email
+
+See an [official spec](#official-specs)'s documentation for spec-specific
+configurations.
+
+## Fetch/Update Archives
+
+When a client has not yet fetched any initial archives, they should clone
+the repository:
+
+    $ git clone https://aur.archlinux.org/archive.git aurweb-archive
+
+When updating, the repository is already cloned and changes need to be pulled
+from remote:
+
+    # To update:
+    $ cd aurweb-archive && git pull
+
+For end-user production applications, see
+[Minimize Disk Space](#minimize-disk-space).
+
+## Minimize Disk Space
+
+Using `git gc` on the repository will compress revisions and remove
+unreachable objects which grow the repository a considerable amount
+each commit. It is recommended that the following command is used
+after cloning the archive or pulling updates:
+
+    $ cd aurweb-archive && git gc --aggressive
+
+## Spec Modules
+
+Each aurweb spec module belongs to the `aurweb.archives.spec` package. For
+example: a spec named "example" would be located at
+`aurweb.archives.spec.example`.
+
+[Official spec listings](#official-specs) use the following format:
+
+- `spec_name`
+    - Spec description; what this spec produces
+        - `<link to repo documentation>`
+
+### Official Specs
+
+- [metadata](doc/specs/metadata.md)
+    - Package RPC `type=info` metadata
+        - [metadata-repo](repos/metadata-repo.md)
+- [users](doc/specs/users.md)
+    - List of users found in the database
+        - [users-repo](repos/users-repo.md)
+- [pkgbases](doc/specs/pkgbases.md)
+    - List of package bases found in the database
+        - [pkgbases-repo](repos/pkgbases-repo.md)
+- [pkgnames](doc/specs/pkgnames.md)
+    - List of package names found in the database
+        - [pkgnames-repo](repos/pkgnames-repo.md)

doc/maintenance.txt

@@ -70,20 +70,48 @@ computations and clean up the database:
 * aurweb-pkgmaint automatically removes empty repositories that were created
   within the last 24 hours but never populated.
 
-* aurweb-mkpkglists generates the package list files; it takes an optional
+* [Deprecated] aurweb-mkpkglists generates the package list files; it takes
-  --extended flag, which additionally produces multiinfo metadata. It also
+  an optional --extended flag, which additionally produces multiinfo metadata.
-  generates {archive.gz}.sha256 files that should be located within
+  It also generates {archive.gz}.sha256 files that should be located within
   mkpkglists.archivedir which contain a SHA-256 hash of their matching
   .gz counterpart.
 
 * aurweb-usermaint removes the last login IP address of all users that did not
   login within the past seven days.
 
+* aurweb-git-archive generates Git repository archives based on a --spec.
+  This script is a new generation of aurweb-mkpkglists, which creates and
+  maintains Git repository versions of the archives produced by
+  aurweb-mkpkglists. See doc/git-archive.md for detailed documentation.
+
 These scripts can be installed by running `poetry install` and are
 usually scheduled using Cron. The current setup is:
 
 ----
-*/5 * * * * poetry run aurweb-mkpkglists [--extended]
+# Run aurweb-git-archive --spec metadata directly after
+# aurweb-mkpkglists so that they are executed sequentially, since
+# both scripts are quite heavy. `aurweb-mkpkglists` should be removed
+# from here once its deprecation period has ended.
+*/5 * * * * poetry run aurweb-mkpkglists [--extended] && poetry run aurweb-git-archive --spec metadata
+
+# Update popularity once an hour. This is done to reduce the amount
+# of changes caused by popularity data. Even if a package is otherwise
+# unchanged, popularity is recalculated every 5 minutes via aurweb-popupdate,
+# which causes changes for a large chunk of packages.
+#
+# At this interval, clients can still take advantage of popularity
+# data, but its updates are guarded behind hour-long intervals.
+*/60 * * * * poetry run aurweb-git-archive --spec popularity
+
+# Usernames
+*/5 * * * * poetry run aurweb-git-archive --spec users
+
+# Package base names
+*/5 * * * * poetry run aurweb-git-archive --spec pkgbases
+
+# Package names
+*/5 * * * * poetry run aurweb-git-archive --spec pkgnames
+
 1 */2 * * * poetry run aurweb-popupdate
 2 */2 * * * poetry run aurweb-aurblup
 3 */2 * * * poetry run aurweb-pkgmaint

doc/repos/metadata-repo.md

@@ -0,0 +1,121 @@
+# Repository: metadata-repo
+
+## Overview
+
+The resulting repository contains RPC `type=info` JSON data for packages,
+split into two different files:
+
+- `pkgbase.json` contains details about each package base in the AUR
+- `pkgname.json` contains details about each package in the AUR
+
+See [Data](#data) for a breakdown of how data is presented in this
+repository based off of a RPC `type=info` base.
+
+See [File Layout](#file-layout) for a detailed summary of the layout
+of these files and the data contained within.
+
+**NOTE: `Popularity` now requires a client-side calculation, see [Popularity Calculation](#popularity-calculation).**
+
+## Data
+
+This repository contains RPC `type=info` data for all packages found
+in AUR's database, reorganized to be suitable for Git repository
+changes.
+
+- `pkgname.json` holds Package-specific metadata
+    - Some fields have been removed from `pkgname.json` objects
+        - `ID`
+        - `PackageBaseID -> ID` (moved to `pkgbase.json`)
+        - `NumVotes` (moved to `pkgbase.json`)
+        - `Popularity` (moved to `pkgbase.json`)
+- `pkgbase.json` holds PackageBase-specific metadata
+    - Package Base fields from `pkgname.json` have been moved over to
+      `pkgbase.json`
+        - `ID`
+        - `Keywords`
+        - `FirstSubmitted`
+        - `LastModified`
+        - `OutOfDate`
+        - `Maintainer`
+        - `URLPath`
+        - `NumVotes`
+        - `Popularity`
+        - `PopularityUpdated`
+
+## Popularity Calculation
+
+Clients intending to use popularity data from this archive **must**
+perform a decay calculation on their end to reflect a close approximation
+of up-to-date popularity.
+
+Putting this step onto the client allows the server to maintain
+less popularity record updates, dramatically improving archiving
+of popularity data. The same calculation is done on the server-side
+when producing outputs for RPC `type=info` and package pages.
+
+```
+Let T = Current UTC timestamp in seconds
+Let PU = PopularityUpdated timestamp in seconds
+
+# The delta between now and PU in days
+Let D = (T - PU) / 86400
+
+# Calculate up-to-date popularity:
+P = Popularity * (0.98^D)
+```
+
+We can see that the resulting up-to-date popularity value decays as
+the exponent is increased:
+- `1.0 * (0.98^1) = 0.98`
+- `1.0 * (0.98^2) = 0.96039999`
+- ...
+
+This decay calculation is essentially pushing back the date found for
+votes by the exponent, which takes into account the time-factor. However,
+since this calculation is based off of decimals and exponents, it
+eventually becomes imprecise. The AUR updates these records on a forced
+interval and whenever a vote is added to or removed from a particular package
+to avoid imprecision from being an issue for clients
+
+## File Layout
+
+#### pkgbase.json:
+
+    {
+        "pkgbase1": {
+            "FirstSubmitted": 123456,
+            "ID": 1,
+            "LastModified": 123456,
+            "Maintainer": "kevr",
+            "OutOfDate": null,
+            "URLPath": "/cgit/aur.git/snapshot/pkgbase1.tar.gz",
+            "NumVotes": 1,
+            "Popularity": 1.0,
+            "PopularityUpdated": 12345567753.0
+        },
+        ...
+    }
+
+#### pkgname.json:
+
+    {
+        "pkg1": {
+            "CheckDepends": [], # Only included if a check dependency exists
+            "Conflicts": [],    # Only included if a conflict exists
+            "Depends": [],      # Only included if a dependency exists
+            "Description": "some description",
+            "Groups": [],       # Only included if a group exists
+            "ID": 1,
+            "Keywords": [],
+            "License": [],
+            "MakeDepends": [],  # Only included if a make dependency exists
+            "Name": "pkg1",
+            "OptDepends": [],   # Only included if an opt dependency exists
+            "PackageBase": "pkgbase1",
+            "Provides": [],     # Only included if `provides` is defined
+            "Replaces": [],     # Only included if `replaces` is defined
+            "URL": "https://some_url.com",
+            "Version": "1.0-1"
+        },
+        ...
+    }

doc/repos/pkgbases-repo.md

@@ -0,0 +1,15 @@
+# Repository: pkgbases-repo
+
+## Overview
+
+- `pkgbase.json` contains a list of package base names
+
+## File Layout
+
+### pkgbase.json:
+
+    [
+        "pkgbase1",
+        "pkgbase2",
+        ...
+    ]

doc/repos/pkgnames-repo.md

@@ -0,0 +1,15 @@
+# Repository: pkgnames-repo
+
+## Overview
+
+- `pkgname.json` contains a list of package names
+
+## File Layout
+
+### pkgname.json:
+
+    [
+        "pkgname1",
+        "pkgname2",
+        ...
+    ]

doc/repos/users-repo.md

@@ -0,0 +1,15 @@
+# Repository: users-repo
+
+## Overview
+
+- `users.json` contains a list of usernames
+
+## File Layout
+
+### users.json:
+
+    [
+        "user1",
+        "user2",
+        ...
+    ]

doc/specs/metadata.md

@@ -0,0 +1,14 @@
+# Git Archive Spec: metadata
+
+## Configuration
+
+- `[git-archive]`
+    - `metadata-repo`
+        - Path to package metadata git repository location
+
+## Repositories
+
+For documentation on each one of these repositories, follow their link,
+which brings you to a topical markdown for that repository.
+
+- [metadata-repo](doc/repos/metadata-repo.md)

doc/specs/pkgbases.md

@@ -0,0 +1,14 @@
+# Git Archive Spec: pkgbases
+
+## Configuration
+
+- `[git-archive]`
+    - `pkgbases-repo`
+        - Path to pkgbases git repository location
+
+## Repositories
+
+For documentation on each one of these repositories, follow their link,
+which brings you to a topical markdown for that repository.
+
+- [pkgbases-repo](doc/repos/pkgbases-repo.md)

doc/specs/pkgnames.md

@@ -0,0 +1,14 @@
+# Git Archive Spec: pkgnames
+
+## Configuration
+
+- `[git-archive]`
+    - `pkgnames-repo`
+        - Path to pkgnames git repository location
+
+## Repositories
+
+For documentation on each one of these repositories, follow their link,
+which brings you to a topical markdown for that repository.
+
+- [pkgnames-repo](doc/repos/pkgnames-repo.md)

doc/specs/popularity.md

@@ -0,0 +1,14 @@
+# Git Archive Spec: popularity
+
+## Configuration
+
+- `[git-archive]`
+    - `popularity-repo`
+        - Path to popularity git repository location
+
+## Repositories
+
+For documentation on each one of these repositories, follow their link,
+which brings you to a topical markdown for that repository.
+
+- [popularity-repo](doc/repos/popularity-repo.md)

doc/specs/users.md

@@ -0,0 +1,14 @@
+# Git Archive Spec: users
+
+## Configuration
+
+- `[git-archive]`
+    - `users-repo`
+        - Path to users git repository location
+
+## Repositories
+
+For documentation on each one of these repositories, follow their link,
+which brings you to a topical markdown for that repository.
+
+- [users-repo](doc/repos/users-repo.md)

migrations/versions/6441d3b65270_add_popularityupdated_to_packagebase.py

@@ -0,0 +1,33 @@
+"""add PopularityUpdated to PackageBase
+
+Revision ID: 6441d3b65270
+Revises: d64e5571bc8d
+Create Date: 2022-09-22 18:08:03.280664
+
+"""
+from alembic import op
+from sqlalchemy.exc import OperationalError
+
+from aurweb.models.package_base import PackageBase
+from aurweb.scripts import popupdate
+
+# revision identifiers, used by Alembic.
+revision = "6441d3b65270"
+down_revision = "d64e5571bc8d"
+branch_labels = None
+depends_on = None
+
+table = PackageBase.__table__
+
+
+def upgrade():
+    try:
+        op.add_column(table.name, table.c.PopularityUpdated)
+    except OperationalError:
+        print(f"table '{table.name}' already exists, skipping migration")
+
+    popupdate.run_variable()
+
+
+def downgrade():
+    op.drop_column(table.name, "PopularityUpdated")

pyproject.toml

@@ -117,3 +117,4 @@ aurweb-tuvotereminder = "aurweb.scripts.tuvotereminder:main"
 aurweb-usermaint = "aurweb.scripts.usermaint:main"
 aurweb-config = "aurweb.scripts.config:main"
 aurweb-adduser = "aurweb.scripts.adduser:main"
+aurweb-git-archive = "aurweb.scripts.git_archive:main"

templates/partials/packages/details.html

@@ -149,7 +149,7 @@
     </tr>
     <tr>
         <th>{{ "Popularity" | tr }}:</th>
-        <td>{{ pkgbase.Popularity | number_format(6 if pkgbase.Popularity <= 0.2 else 2) }}</td>
+        <td>{{ popularity | number_format(6 if popularity <= 0.2 else 2) }}</td>
     </tr>
     <tr>
         <th>{{ "First Submitted" | tr }}:</th>

test/test_git_archives.py

@@ -0,0 +1,241 @@
+from http import HTTPStatus
+from typing import Tuple
+from unittest import mock
+
+import py
+import pygit2
+import pytest
+from fastapi.testclient import TestClient
+
+from aurweb import asgi, config, db
+from aurweb.archives.spec.base import GitInfo, SpecBase
+from aurweb.models import Package, PackageBase, User
+from aurweb.scripts import git_archive
+from aurweb.testing.requests import Request
+
+
+@pytest.fixture
+def mock_metadata_archive(
+    tmp_path: py.path.local,
+) -> Tuple[py.path.local, py.path.local]:
+    metadata_path = tmp_path / "metadata.git"
+
+    get_ = config.get
+
+    def mock_config(section: str, option: str) -> str:
+        if section == "git-archive":
+            if option == "metadata-repo":
+                return str(metadata_path)
+        return get_(section, option)
+
+    with mock.patch("aurweb.config.get", side_effect=mock_config):
+        yield metadata_path
+
+
+@pytest.fixture
+def mock_users_archive(tmp_path: py.path.local) -> py.path.local:
+    users_path = tmp_path / "users.git"
+
+    get_ = config.get
+
+    def mock_config(section: str, option: str) -> str:
+        if section == "git-archive":
+            if option == "users-repo":
+                return str(users_path)
+        return get_(section, option)
+
+    with mock.patch("aurweb.config.get", side_effect=mock_config):
+        yield users_path
+
+
+@pytest.fixture
+def mock_pkgbases_archive(tmp_path: py.path.local) -> py.path.local:
+    pkgbases_path = tmp_path / "pkgbases.git"
+
+    get_ = config.get
+
+    def mock_config(section: str, option: str) -> str:
+        if section == "git-archive":
+            if option == "pkgbases-repo":
+                return str(pkgbases_path)
+        return get_(section, option)
+
+    with mock.patch("aurweb.config.get", side_effect=mock_config):
+        yield pkgbases_path
+
+
+@pytest.fixture
+def mock_pkgnames_archive(tmp_path: py.path.local) -> py.path.local:
+    pkgnames_path = tmp_path / "pkgnames.git"
+
+    get_ = config.get
+
+    def mock_config(section: str, option: str) -> str:
+        if section == "git-archive":
+            if option == "pkgnames-repo":
+                return str(pkgnames_path)
+        return get_(section, option)
+
+    with mock.patch("aurweb.config.get", side_effect=mock_config):
+        yield pkgnames_path
+
+
+@pytest.fixture
+def metadata(mock_metadata_archive: py.path.local) -> py.path.local:
+    args = [__name__, "--spec", "metadata"]
+    with mock.patch("sys.argv", args):
+        yield mock_metadata_archive
+
+
+@pytest.fixture
+def users(mock_users_archive: py.path.local) -> py.path.local:
+    args = [__name__, "--spec", "users"]
+    with mock.patch("sys.argv", args):
+        yield mock_users_archive
+
+
+@pytest.fixture
+def pkgbases(mock_pkgbases_archive: py.path.local) -> py.path.local:
+    args = [__name__, "--spec", "pkgbases"]
+    with mock.patch("sys.argv", args):
+        yield mock_pkgbases_archive
+
+
+@pytest.fixture
+def pkgnames(mock_pkgnames_archive: py.path.local) -> py.path.local:
+    args = [__name__, "--spec", "pkgnames"]
+    with mock.patch("sys.argv", args):
+        yield mock_pkgnames_archive
+
+
+@pytest.fixture
+def client() -> TestClient:
+    yield TestClient(app=asgi.app)
+
+
+@pytest.fixture
+def user(db_test: None) -> User:
+    with db.begin():
+        user_ = db.create(
+            User,
+            Username="test",
+            Email="test@example.org",
+            Passwd="testPassword",
+        )
+
+    yield user_
+
+
+@pytest.fixture
+def package(user: User) -> Package:
+    with db.begin():
+        pkgbase_ = db.create(
+            PackageBase,
+            Name="test",
+            Maintainer=user,
+            Packager=user,
+        )
+
+        pkg_ = db.create(
+            Package,
+            PackageBase=pkgbase_,
+            Name="test",
+        )
+
+    yield pkg_
+
+
+def commit_count(repo: pygit2.Repository) -> int:
+    commits = 0
+    for _ in repo.walk(repo.head.target):
+        commits += 1
+    return commits
+
+
+def test_specbase_raises_notimplementederror():
+    spec = SpecBase()
+    with pytest.raises(NotImplementedError):
+        spec.generate()
+
+
+def test_gitinfo_config(tmpdir: py.path.local):
+    path = tmpdir / "test.git"
+    git_info = GitInfo(path, {"user.name": "Test Person"})
+    git_archive.init_repository(git_info)
+
+    repo = pygit2.Repository(path)
+    assert repo.config["user.name"] == "Test Person"
+
+
+def test_metadata(metadata: py.path.local, package: Package):
+    # Run main(), which creates mock_metadata_archive and commits current
+    # package data to it, exercising the "diff detected, committing" path
+    assert git_archive.main() == 0
+    repo = pygit2.Repository(metadata)
+    assert commit_count(repo) == 1
+
+    # Run main() again to exercise the "no diff detected" path
+    assert git_archive.main() == 0
+    repo = pygit2.Repository(metadata)
+    assert commit_count(repo) == 1
+
+
+def test_metadata_change(
+    client: TestClient, metadata: py.path.local, user: User, package: Package
+):
+    """Test that metadata changes via aurweb cause git_archive to produce diffs."""
+    # Run main(), which creates mock_metadata_archive and commits current
+    # package data to it, exercising the "diff detected, committing" path
+    assert git_archive.main() == 0
+    repo = pygit2.Repository(metadata)
+    assert commit_count(repo) == 1
+
+    # Now, we modify `package`-related metadata via aurweb POST.
+    pkgbasename = package.PackageBase.Name
+    cookies = {"AURSID": user.login(Request(), "testPassword")}
+
+    with client as request:
+        endp = f"/pkgbase/{pkgbasename}/keywords"
+        post_data = {"keywords": "abc def"}
+        resp = request.post(endp, data=post_data, cookies=cookies, allow_redirects=True)
+    assert resp.status_code == HTTPStatus.OK
+
+    # Run main() again, which should now produce a new commit with the
+    # keyword changes we just made
+    assert git_archive.main() == 0
+    repo = pygit2.Repository(metadata)
+    assert commit_count(repo) == 2
+
+
+def test_metadata_delete(client: TestClient, metadata: py.path.local, package: Package):
+    # Run main(), which creates mock_metadata_archive and commits current
+    # package data to it, exercising the "diff detected, committing" path
+    assert git_archive.main() == 0
+    repo = pygit2.Repository(metadata)
+    assert commit_count(repo) == 1
+
+    with db.begin():
+        db.delete(package)
+
+    # The deletion here should have caused a diff to be produced in git
+    assert git_archive.main() == 0
+    repo = pygit2.Repository(metadata)
+    assert commit_count(repo) == 2
+
+
+def test_users(users: py.path.local, user: User):
+    assert git_archive.main() == 0
+    repo = pygit2.Repository(users)
+    assert commit_count(repo) == 1
+
+
+def test_pkgbases(pkgbases: py.path.local, package: Package):
+    assert git_archive.main() == 0
+    repo = pygit2.Repository(pkgbases)
+    assert commit_count(repo) == 1
+
+
+def test_pkgnames(pkgnames: py.path.local, package: Package):
+    assert git_archive.main() == 0
+    repo = pygit2.Repository(pkgnames)
+    assert commit_count(repo) == 1

test/test_templates.py

@@ -9,6 +9,7 @@ from aurweb.filters import as_timezone, number_format, timestamp_to_datetime as
 from aurweb.models import Package, PackageBase, User
 from aurweb.models.account_type import USER_ID
 from aurweb.models.license import License
+from aurweb.models.package_base import popularity
 from aurweb.models.package_license import PackageLicense
 from aurweb.models.package_relation import PackageRelation
 from aurweb.models.relation_type import PROVIDES_ID, REPLACES_ID
@@ -287,12 +288,14 @@ def test_package_details(user: User, package: Package):
     """Test package details with most fields populated, but not all."""
     request = Request(user=user, authenticated=True)
     context = make_context(request, "Test Details")
+
     context.update(
         {
             "request": request,
             "git_clone_uri_anon": GIT_CLONE_URI_ANON,
             "git_clone_uri_priv": GIT_CLONE_URI_PRIV,
             "pkgbase": package.PackageBase,
+            "popularity": popularity(package.PackageBase, time.utcnow()),
             "package": package,
             "comaintainers": [],
         }
@@ -329,6 +332,7 @@ def test_package_details_filled(user: User, package: Package):
             "git_clone_uri_anon": GIT_CLONE_URI_ANON,
             "git_clone_uri_priv": GIT_CLONE_URI_PRIV,
             "pkgbase": package.PackageBase,
+            "popularity": popularity(package.PackageBase, time.utcnow()),
             "package": package,
             "comaintainers": [],
             "licenses": package.package_licenses,