Improve transaction list and report performance with four targeted query and index changes

Open
#466 0 comments 0 reactions 0 assignees View on GitHub

Nobody has claimed this yet.

Assessment

Difficulty
4/5
Estimated time
3-5 days
Newbie friendliness
35/100
Issue type
Bug
Clarity
Mostly clear
Activity status
Active
Tech stack
csharp

Research direction

Start with the list query in TransactionsController.cs, the report query in ReportsController.cs, and the report projection in TransfersReportViewModel.cs; read the shared _TransactionsTable.cshtml before changing its models or partials. Compare generated SQL and behavior for transactions with multiple transfers and metadata, then measure the proposed projections and covering index using the listed TEST ranges. Done means preserving filters and report contents while recording performance results and a reversible index migration plan.

Written by the indexing model from the issue text.

Description

Larger date ranges make the transaction list and downloadable transactions report slow or unusable on the DEVAR production instance. Reduce database reads, duplicated result rows, and application memory use through four targeted changes, preserving the existing filters, displayed values, and exports.

Read-only investigation on October 21, 2026 found:

  • DEVAR has one team and one source, approximately 275,634 transactions, 1.16 million transfers, and 803,000 transaction metadata records. Its default date range is already seven days.
  • Over the preceding 14 days, the main transaction list query averaged 13.4 seconds across 171 completed executions, peaked at 176 seconds, and recorded two aborted executions around 30 seconds.
  • The downloadable report recorded a 26-second completion and two aborted executions, one at 30 seconds. The completed execution returned 33,267 joined rows and performed approximately 494,450 logical reads.
  • Existing transaction-date and transfer-transaction indexes are present and used by the execution plans. Both queries still perform many lookups into full rows. Query Store shows substantial data-read waits, and Azure metrics show read-I/O utilization reaching 100%.
  • The list loads every matching transaction and its transfers before rendering HTML. The report loads transfers and metadata and renders one table row per transfer. DataTables paginates the resulting HTML in the browser.

Implement and evaluate these four changes:

  1. Split the downloadable report query and disable tracking. Add query-local AsSplitQuery() and AsNoTracking(). The current generated SQL joins both transfers and metadata at the transaction level: four transfers and three metadata entries produce 12 intermediate rows. Fetching the collections separately should remove this multiplication. Keep this scoped to the report rather than changing the DbContext default. Verify behavior if transactions change between the separate reads.

  2. Calculate report totals and transfer counts once from the transfers already loaded. The current projection loads transfers and additionally issues correlated SUM and COUNT subqueries over them. The stored plan shows repeated transfer reads and lookups. After materialization, calculate the credit-only total and transfer count from the loaded collection, preserving decimal precision and existing report semantics.

  3. Project only the fields needed by the list and report. Both paths currently retrieve full transfer entities, including unused legacy accounting fields. Introduce focused read models for the displayed transaction fields and required transfer fields. Preserve the transaction-description/debit-description fallback, credit totals, account strings, metadata formatting, transfer ordering, links, and hidden fields used by search or exports. Check other consumers before changing shared transaction-table models or partials.

  4. Benchmark a covering transfer index alongside the smaller projections. Candidate: Transfers(TransactionId) INCLUDE (Direction, Amount, Description, FinancialSegmentString). The existing IX_Transfers_TransactionId does not contain these displayed fields, so the current plans fetch them through additional table lookups. Confirm the final projected columns and execution plans before choosing the index definition. Measure read benefits, index size, and write overhead; avoid introducing a redundant index without evidence. A source/date composite index is lower priority for DEVAR because its single source adds little selectivity and the date index is already used.

Relevant code:

Validate in TEST with representative DEVAR-scale data:

  • Compare before/after query time, logical reads, returned SQL rows, and application memory for identical 7-, 30-, and 90-day ranges. Record first-observed and repeated timings separately.
  • Confirm identical transaction and transfer counts, credit totals, descriptions, accounts, metadata, and export contents. Include transactions with multiple transfers and metadata entries.
  • Preserve team authorization, tracking-number lookup behavior, merchant filters, and date-range boundaries.
  • Verify generated SQL avoids the transfers/metadata cross product and repeated report aggregate subqueries, and confirm the candidate index helps the final query shape before retaining it.
  • Provide a reversible index migration and rollback plan. Production application or schema changes require separate approval after TEST validation.

These are evidence-backed candidates, not measured improvements yet. Server-side pagination, server-generated exports, and Azure tier upgrades are separate work. The existing unbounded HTML response remains a scaling limit after these optimizations. This issue records the investigation and proposed scope; no implementation or production changes have been made.

Dominant language
C#
Stars
4
Forks
2
PR merge metrics
No merged PRs in 30d

Getting set up

This project ships no dev container, Dockerfile or contributing guide, so setting up is up to you: start from its README, and see our first-contribution guide for the general steps.

First steps

  1. Read the whole issue, then the project's contributing guide.
  2. Comment on the issue to say you are picking it up — it saves two people doing the same work.
  3. Fork the repository and make your change on a branch.
  4. Open a pull request that references the issue number.

More from ucdavis/Sloth

All issues in ucdavis/Sloth

Similar issues

More C# issues

Get new issues in your inbox

A short digest of beginner-friendly GitHub issues.