pst

Read and edit in place Microsoft PST files with rich schemas for common MAPI types (emails, contacts, appointments, tasks)

Maintainer(s): mach-kernel

Installing and Loading

INSTALL pst FROM community;
LOAD pst;

Example

-- Count all messages across PST files (supports globs)
D SELECT count(*) FROM read_pst_messages('enron/*.pst');
┌────────────────┐
│  count_star()  │
│     int64      │
├────────────────┤
│    1227193     │
└────────────────┘

-- Query contacts (supports remote URIs)
D SELECT given_name, surname FROM read_pst_contacts('https://example.com/outlook.pst');
┌────────────┬─────────┐
│ given_name │ surname │
│  varchar   │ varchar │
├────────────┼─────────┤
│ John       │ Doe     │
│ Jane       │ Smith   │
└────────────┴─────────┘

-- Read messages with limit (applied during planning for large files)
D SELECT subject, sender_email_address, message_delivery_time
  FROM read_pst_messages('*.pst', read_limit=100);

-- Preview an in-place delete. Without `really := true` the call
-- opens read-only and reports what it would do per row.
D SELECT status, count(*) FROM delete_pst_messages(
    SELECT pst_path, node_id FROM read_pst_messages('/tmp/hr.pst')
    WHERE sender_email_address = '[email protected]'
  ) GROUP BY status;
┌───────────┬──────────────┐
│  status   │ count_star() │
│  varchar  │    int64     │
├───────────┼──────────────┤
│ PREVIEWED │           17 │
└───────────┴──────────────┘

About pst

A DuckDB extension for reading and editing in place Microsoft PST files, with rich schemas for common MAPI types. Built on Microsoft’s official PST SDK. Query emails, contacts, appointments, and more. Use it to analyze PST data in-place (locally or on object storage), import to DuckDB tables, export to Parquet, or delete from a store in place to scrub PII, drop out-of-scope custodian mail before a production, or strip attachments.

Table Functions

In-place Deletion

Gated behind SET pst_allow_delete = true (session) plus really := true per call. Without really the call opens read-only, resolves every target, and reports PREVIEWED or FAILED per row. Bind errors abort the statement; anything found later is reported per row so a mid-statement throw cannot leave already-committed deletes with no record of what went. Node IDs are checked against the mode, so delete_pst_messages refuses a folder or the store’s root.

Performance Features

• Query pushdown: projection and statistics pushdown
• Concurrent planning: parallel partition planning for directories with many PST files
• Late materialization: filter on virtual columns before expanding full projections

Parameters

For full schema documentation and usage examples, see the GitHub repository.

Added Functions

Overloaded Functions

This extension does not add any function overloads.

Added Types

This extension does not add any types.

Added Settings