Eliminate N+1 queries in scheduled tasks

[Copilot]

Both sync_emails() and enable_pubsub_everyday() fetched account names with get_all() then immediately loaded full documents with get_doc() in a loop. With sync_emails() running every 30 minutes, this creates unnecessary database load.

Changes

sync.py - Eliminated get_doc() entirely

# Before: Only fetch name, then get_doc() for each account
gmail_accounts = frappe.get_all("Gmail Account", filters={"gmail_enabled": 1}, fields=["name"])
for gmail_account in gmail_accounts:
    gaccount = frappe.get_doc("Gmail Account", gmail_account.name)
    if gaccount.refresh_token:
        user = gaccount.linked_user

# After: Fetch needed fields directly
gmail_accounts = frappe.get_all("Gmail Account", filters={"gmail_enabled": 1}, 
                                fields=["name", "refresh_token", "linked_user"])
for gmail_account in gmail_accounts:
    if gmail_account.refresh_token:
        user = gmail_account.linked_user

daily.py - Filter at query time instead of in Python

# Before: Load all enabled accounts
filters={"gmail_enabled": 1}

# After: Filter accounts without tokens upfront
filters={"gmail_enabled": 1, "refresh_token": ["is", "set"]}

Note: daily.py still requires get_doc() since enable_pubsub() accesses the labels child table, but the filter reduces unnecessary iterations.

Original prompt

---

This section details on the original issue you should resolve

<issue_title>get_doc() in Loop for Sync and Daily Tasks</issue_title>
<issue_description>

Metadata

• File(s):
  • frappe_gmail_thread/tasks/sync.py:12
  • frappe_gmail_thread/tasks/daily.py:20
• Category: Database / Scheduled Tasks
• Severity: Medium
• Effort to Fix: Low
• Estimated Performance Gain: 20-40%

Problem Description

Both scheduled tasks (sync_emails and enable_pubsub_everyday) fetch Gmail Account names with get_all(), then immediately load the full document with get_doc() in a loop. This creates unnecessary database queries when only specific fields are needed.

The sync_emails task runs every 30 minutes, and enable_pubsub_everyday runs daily. With many Gmail accounts, these N+1 query patterns add up.

Code Location

sync.py (line 12):

def sync_emails():
    gmail_accounts = frappe.get_all(
        "Gmail Account",
        filters={"gmail_enabled": 1},
        fields=["name"],  # Only fetches name
    )
    for gmail_account in gmail_accounts:
        gaccount = frappe.get_doc("Gmail Account", gmail_account.name)  # Full doc load
        if gaccount.refresh_token:
            user = gaccount.linked_user
            # ...

daily.py (line 20):

def enable_pubsub_everyday():
    # ...
    gmail_accounts = frappe.get_all(
        "Gmail Account", filters={"gmail_enabled": 1}, fields=["name"]
    )
    for gmail_account in gmail_accounts:
        gaccount = frappe.get_doc("Gmail Account", gmail_account.name)  # Full doc load
        try:
            enable_pubsub(gaccount)
        # ...

Root Cause

The pattern of:

1. Fetch list with minimal fields
2. Loop and load full document for each item

This is inefficient because:

• Initial query returns only name
• Each get_doc() requires a separate database round-trip
• Full document loads all fields, child tables, and runs controller logic

Proposed Solution

For sync.py - Only need refresh_token and linked_user:

def sync_emails():
    gmail_accounts = frappe.get_all(
        "Gmail Account",
        filters={"gmail_enabled": 1},
        fields=["name", "refresh_token", "linked_user"],  # Fetch needed fields
    )
    for gmail_account in gmail_accounts:
        if gmail_account.refresh_token:
            user = gmail_account.linked_user
            job_name = f"gmail_thread_sync_{user}"
            if not is_job_enqueued(job_name):
                frappe.enqueue(
                    "frappe_gmail_thread.frappe_gmail_thread.doctype.gmail_thread.gmail_thread.sync",
                    user=user,
                    queue="long",
                    job_name=job_name,
                    job_id=job_name,
                )

For daily.py - enable_pubsub needs full doc, but can batch check first:

If enable_pubsub() truly needs the full document, at minimum filter accounts that have the necessary tokens:

def enable_pubsub_everyday():
    google_settings = frappe.get_single("Google Settings")
    if not google_settings.enable:
        return
    if (
        not google_settings.custom_gmail_sync_in_realtime
        or not google_settings.custom_gmail_pubsub_topic
    ):
        return

    # Only fetch accounts with refresh tokens (they're the only ones that can use pubsub)
    gmail_accounts = frappe.get_all(
        "Gmail Account", 
        filters={
            "gmail_enabled": 1,
            "refresh_token": ["is", "set"]  # Filter out accounts without tokens
        }, 
        fields=["name"]
    )
    for gmail_account in gmail_accounts:
        gaccount = frappe.get_doc("Gmail Account", gmail_account.name)
        try:
            enable_pubsub(gaccount)
        except Exception as e:
            frappe.log_error(title="PubSub Error", message=e)
            continue

Implementation Steps

1. sync.py:

   • Update fields to include ["name", "refresh_token", "linked_user"]
   • Remove get_doc() call
   • Access fields directly from the dict returned by get_all()

2. daily.py:

   • Add filter for refresh_token to reduce iterations
   • If enable_pubsub() can be refactored to accept specific fields, do so
   • Otherwise, the filter optimization still reduces unnecessary iterations

3. Test both scheduled tasks with multiple Gmail accounts

Additional Notes

• The sync() function called by enqueue() does need the full document, but that's inside the background job
• The scheduler only needs to check if tokens exist and get the user email
• This pattern appears across multiple rtCamp repos (frappe-slack-connector, frappe-github-connector)</issue_description>

Comments on the Issue (you are @copilot in this section)

• Fixes get_doc() in Loop for Sync and Daily Tasks #45

---

💬 We'd love your input! Share your thoughts on Copilot coding agent in our 2 minute survey.