mkpkglists improvements

This overlaid MR changes a couple things, namely the format of the meta package archive and only updating archives when needed.

Update When Needed:

We now use the information_schema.tables AUTO_INCREMENT column to keep track of a "serial number" that can be used to tell us whether we need to update or not.

Metadata Format:

    The SQL logic in this file for package metadata now exactly
    reflects RPC's search logic, without searching for specific
    packages.

    Two command line arguments are available:

        --extended | Include License, Keywords and Groups fields

    When one of these arguments are passed, the script will create
    a packages-meta-ext-v1.json.gz, configured via packagesmetaextfile.

    Archive JSON is in the following format: line-separated package objects
    enclosed in a list:

        [
        {...},
        {...},
        {...}
        ]

    This allows users to stream the archive and process packages
    line-by-line, during the local storage process.

    Signed-off-by: Kevin Morris <kevr@0cost.org>

aurweb/scripts/mkpkglists.py

 #!/usr/bin/env python3
+"""
+Produces package, package base and user archives for the AUR
+database.
+
+Archives:
+
+    packages.gz               | A line-separated list of package names
+    packages-meta-v1.json     | A type=search RPC-formatted JSON dataset
+    packages-meta-ext-v1.json | An --extended or --all archive
+    pkgbase.gz                | A line-separated list of package base names
+    users.gz                  | A line-separated list of user names
+
+This script takes an optional argument: --extended or --all. Based
+on the following, right-hand side fields are added to each item.
+
+    --extended  | License, Keywords, Groups
+    --all       | License, Keywords, Groups, relations and dependencies
+
+"""
 
 import datetime
 import gzip
-import json
 import os
+import sys
+
+from collections import defaultdict
+from decimal import Decimal
+from typing import Tuple
+
+import orjson
 
 import aurweb.config
 import aurweb.db
 
+
+def state_path(archive: str) -> str:
+    # A hard-coded /tmp state directory.
+    # TODO: Use Redis cache to store this state after we merge
+    # FastAPI into master and removed PHP from the tree.
+    return os.path.join("/tmp", os.path.basename(archive) + ".state")
+
+
 packagesfile = aurweb.config.get('mkpkglists', 'packagesfile')
 packagesmetafile = aurweb.config.get('mkpkglists', 'packagesmetafile')
+packagesmetaextfile = aurweb.config.get('mkpkglists', 'packagesmetaextfile')
+packages_state = state_path(packagesfile)
+
 pkgbasefile = aurweb.config.get('mkpkglists', 'pkgbasefile')
+pkgbases_state = state_path(pkgbasefile)
+
 userfile = aurweb.config.get('mkpkglists', 'userfile')
+users_state = state_path(userfile)
 
 
-def should_update(tablename: str) -> int:
+def should_update(state: str, tablename: str) -> Tuple[bool, int]:
     if aurweb.config.get("database", "backend") != "mysql":
-        return False
-
-    conn = aurweb.db.Connection()
+        return (False, 0)
 
     db_name = aurweb.config.get("database", "name")
+    conn = aurweb.db.Connection()
     cur = conn.execute("SELECT auto_increment FROM information_schema.tables "
                        "WHERE table_schema = ? AND table_name = ?",
                        (db_name, tablename,))
     update_time = cur.fetchone()[0]
 
-    cached_update_time = 0
-    if os.path.exists(f"/tmp/{tablename}.update-time.cache"):
-        with open(f"/tmp/{tablename}.update-time.cache") as f:
-            cached_update_time = int(f.read().strip())
+    saved_update_time = 0
+    if os.path.exists(state):
+        with open(state) as f:
+            saved_update_time = int(f.read().strip())
 
-    return (cached_update_time == update_time, update_time)
+    return (saved_update_time == update_time, update_time)
 
 
-def update_cache(tablename: str, update_time: int) -> None:
-    with open(f"/tmp/{tablename}.update-time.cache", "w") as f:
+def update_state(state: str, update_time: int) -> None:
+    with open(state, "w") as f:
         f.write(str(update_time))
 
 
+TYPE_MAP = {
+    "depends": "Depends",
+    "makedepends": "MakeDepends",
+    "checkdepends": "CheckDepends",
+    "optdepends": "OptDepends",
+    "conflicts": "Conflicts",
+    "provides": "Provides",
+    "replaces": "Replaces",
+}
+
+
+def get_extended_dict(query: str):
+    """
+    Produce data in the form in a single bulk SQL query:
+
+    {
+        <integer_package_id>: {
+            "Depends": [...],
+            "Conflicts": [...],
+            "License": [...]
+        }
+    }
+
+    The caller can then use this data to populate a dataset of packages.
+
+    output = produce_base_output_data()
+    data = get_extended_dict(query)
+    for i in range(len(output)):
+        package_id = output[i].get("ID")
+        output[i].update(data.get(package_id))
+    """
+
+    conn = aurweb.db.Connection()
+
+    cursor = conn.execute(query)
+
+    data = defaultdict(lambda: defaultdict(list))
+
+    for result in cursor.fetchall():
+
+        pkgid = result[0]
+        key = TYPE_MAP.get(result[1])
+        output = result[2]
+        if result[3]:
+            output += result[3]
+
+        # In all cases, we have at least an empty License list.
+        if "License" not in data[pkgid]:
+            data[pkgid]["License"] = []
+
+        # In all cases, we have at least an empty Keywords list.
+        if "Keywords" not in data[pkgid]:
+            data[pkgid]["Keywords"] = []
+
+        if "Groups" not in data[pkgid]:
+            data[pkgid]["Groups"] = []
+
+        data[pkgid][key].append(output)
+
+    conn.close()
+    return data
+
+
+def get_limited_extended_fields():
+    # Returns: [ID, Type, Name, Cond]
+    query = """
+    SELECT PackageGroups.PackageID AS ID, 'Groups' AS Type,
+           Groups.Name, '' AS Cond
+    FROM Groups INNER JOIN PackageGroups
+    ON PackageGroups.GroupID = Groups.ID
+    UNION SELECT PackageLicenses.PackageID AS ID, 'License' AS Type,
+          Licenses.Name, '' as Cond
+    FROM Licenses INNER JOIN PackageLicenses
+    ON PackageLicenses.LicenseID = Licenses.ID
+    UNION SELECT Packages.ID AS ID, 'Keywords' AS Type,
+          PackageKeywords.Keyword AS Name, '' as Cond
+    FROM PackageKeywords
+    INNER JOIN Packages ON Packages.PackageBaseID = PackageKeywords.PackageBaseID
+    """
+    return get_extended_dict(query)
+
+
+def get_extended_fields():
+    # Returns: [ID, Type, Name, Cond]
+    query = """
+    SELECT PackageDepends.PackageID AS ID, DependencyTypes.Name AS Type,
+           PackageDepends.DepName AS Name, PackageDepends.DepCondition AS Cond
+    FROM PackageDepends
+    LEFT JOIN DependencyTypes
+    ON DependencyTypes.ID = PackageDepends.DepTypeID
+    UNION SELECT PackageRelations.PackageID AS ID, RelationTypes.Name AS Type,
+          PackageRelations.RelName AS Name,
+          PackageRelations.RelCondition AS Cond
+    FROM PackageRelations
+    LEFT JOIN RelationTypes
+    ON RelationTypes.ID = PackageRelations.RelTypeID
+    UNION SELECT PackageGroups.PackageID AS ID, 'Groups' AS Type,
+          Groups.Name, '' AS Cond
+    FROM Groups
+    INNER JOIN PackageGroups ON PackageGroups.GroupID = Groups.ID
+    UNION SELECT PackageLicenses.PackageID AS ID, 'License' AS Type,
+          Licenses.Name, '' as Cond
+    FROM Licenses
+    INNER JOIN PackageLicenses ON PackageLicenses.LicenseID = Licenses.ID
+    UNION SELECT Packages.ID AS ID, 'Keywords' AS Type,
+          PackageKeywords.Keyword AS Name, '' as Cond
+    FROM PackageKeywords
+    INNER JOIN Packages ON Packages.PackageBaseID = PackageKeywords.PackageBaseID
+    """
+    return get_extended_dict(query)
+
+
+EXTENDED_FIELD_HANDLERS = {
+    "--extended": get_limited_extended_fields,
+    "--all": get_extended_fields
+}
+
+
+def is_decimal(column):
+    """ Check if an SQL column is of decimal.Decimal type. """
+    if isinstance(column, Decimal):
+        return float(column)
+    return column
+
+
+def write_archive(archive: str, output: list):
+    with gzip.open(archive, "wb") as f:
+        f.write(b"[\n")
+        for i, item in enumerate(output):
+            f.write(orjson.dumps(item))
+            if i < len(output) - 1:
+                f.write(b",")
+            f.write(b"\n")
+        f.write(b"]")
+
+
 def main():
     conn = aurweb.db.Connection()
 
@@ -47,94 +221,87 @@ def main():
     pkgbaselist_header = "# AUR package base list, generated on " + datestr
     userlist_header = "# AUR user name list, generated on " + datestr
 
-    updated, update_time = should_update("Packages")
+    updated, update_time = should_update(packages_state, "Packages")
     if not updated:
         print("Updating Packages...")
-        columns = ("Packages.ID, PackageBaseID, Packages.Name, "
-                   "Version, Description, URL")
+
+        # Query columns; copied from RPC.
+        columns = ("Packages.ID, Packages.Name, "
+                   "PackageBases.ID AS PackageBaseID, "
+                   "PackageBases.Name AS PackageBase, "
+                   "Version, Description, URL, NumVotes, "
+                   "Popularity, OutOfDateTS AS OutOfDate, "
+                   "Users.UserName AS Maintainer, "
+                   "SubmittedTS AS FirstSubmitted, "
+                   "ModifiedTS AS LastModified")
+
+        # Perform query.
         cur = conn.execute(f"SELECT {columns} FROM Packages "
-                           "INNER JOIN PackageBases "
+                           "LEFT JOIN PackageBases "
                            "ON PackageBases.ID = Packages.PackageBaseID "
+                           "LEFT JOIN Users "
+                           "ON PackageBases.MaintainerUID = Users.ID "
                            "WHERE PackageBases.PackagerUID IS NOT NULL")
-        results = cur.fetchall()
 
-        with gzip.open(packagesfile, "w") as f:
+        # Produce packages-meta-v1.json.gz
+        output = list()
+        snapshot_uri = aurweb.config.get("options", "snapshot_uri")
+        for result in cur.fetchall():
+            item = {
+                column[0]: is_decimal(result[i])
+                for i, column in enumerate(cur.description)
+            }
+            item["URLPath"] = snapshot_uri % item.get("Name")
+            output.append(item)
+
+        write_archive(packagesmetafile, output)
+
+        # Produce packages-meta-ext-v1.json.gz
+        if len(sys.argv) > 1 and sys.argv[1] in EXTENDED_FIELD_HANDLERS:
+            f = EXTENDED_FIELD_HANDLERS.get(sys.argv[1])
+            data = f()
+
+            default_ = {"Groups": [], "License": [], "Keywords": []}
+            for i in range(len(output)):
+                data_ = data.get(output[i].get("ID"), default_)
+                output[i].update(data_)
+
+            write_archive(packagesmetaextfile, output)
+
+        # Produce packages.gz
+        with gzip.open(packagesfile, "wb") as f:
             f.write(bytes(pkglist_header + "\n", "UTF-8"))
-            f.writelines([bytes(x[2] + "\n", "UTF-8") for x in results])
-
-        with gzip.open(packagesmetafile, "wt") as f:
-            """ The output "data" json key points to a list of dictionaries,
-            each representing a single result, filled with column names as
-            keys and column values as values.
-
-            The output "mapping" json key points to a dictionary of Package
-            name key -> "data"-list index pairs. This provides users of
-            the meta archive a way to perform O(1) searches based on a
-            package name, while still providing a sequential list for
-            loopability.
-
-            i = json_data["mapping"]["package_name"]
-            package_data = json_data["data"][i]
-
-            name = package_data.get("Name")
-            version = package_data.get("Version")
-
-            Example:
-                {
-                    "data": [
-                        {
-                            "ID": 123,
-                            "Name": "package_name",
-                            "PackageBaseID": 234,
-                            "Version": "0.1.1",
-                            "Description": "Some description...",
-                            "URL": "https://some.url"
-                        },
-                        ...
-                    ],
-                    "mapping": {
-                        "package_name": 0,
-                        ...
-                    }
-                }
-            """
-            json.dump({
-                "warning": ("This is a experimental! It can be removed "
-                            "or modified without warning!"),
-                "mapping": {
-                    result[2]: i
-                    for i, result in enumerate(results)
-                },
-                "data": [{
-                    column[0]: result[i]
-                    for i, column in enumerate(cur.description)
-                } for result in results]
-            }, f)
-
-        update_cache("Packages", update_time)
+            f.writelines([
+                bytes(x.get("Name") + "\n", "UTF-8")
+                for x in output
+            ])
+
+        update_state(packages_state, update_time)
     else:
         print("Packages have not been updated; skipping.")
 
-    updated, update_time = should_update("PackageBases")
+    updated, update_time = should_update(pkgbases_state, "PackageBases")
     if not updated:
         print("Updating PackageBases...")
+        # Produce pkgbase.gz
         with gzip.open(pkgbasefile, "w") as f:
             f.write(bytes(pkgbaselist_header + "\n", "UTF-8"))
             cur = conn.execute("SELECT Name FROM PackageBases " +
                                "WHERE PackagerUID IS NOT NULL")
             f.writelines([bytes(x[0] + "\n", "UTF-8") for x in cur.fetchall()])
-        update_cache("PackageBases", update_time)
+        update_state(pkgbases_state, update_time)
     else:
         print("PackageBases have not been updated; skipping.")
 
-    updated, update_time = should_update("Users")
+    updated, update_time = should_update(users_state, "Users")
     if not updated:
         print("Updating Users...")
+        # Produce users.gz
         with gzip.open(userfile, "w") as f:
             f.write(bytes(userlist_header + "\n", "UTF-8"))
             cur = conn.execute("SELECT UserName FROM Users")
             f.writelines([bytes(x[0] + "\n", "UTF-8") for x in cur.fetchall()])
-        update_cache("Users", update_time)
+        update_state(users_state, update_time)
     else:
         print("Users have not been updated; skipping.")