Reconciliation
Indexed, deterministic work
Parsing, normalization, indexing, matching, and result preparation run locally in dedicated browser workers.
Limits and measured behavior
Clear limits for local spreadsheet reconciliation.
MatchDock's reconciliation engine and its XLSX exporter have different measured costs. This page separates them, states the current safeguards, and identifies exactly which browser environments have been tested.
Large XLSX export
More time and memory
Workbook generation runs in a dedicated export worker, but still requires the full in-memory workbook before the Blob download begins.
Current input caps
These safeguards stop processing before known browser-risk boundaries are exceeded.
- 50 MiB
- per file
- 200,000
- data rows per source
- 500
- columns per source
These parser safeguards are not a promise that every dataset below all three caps will fit available browser memory. Wider rows, large cell values, complex rules, XLSX parsing, and duplicate-heavy results can require more resources.
Reconciliation is not export
Reconciliation
Parsing and reconciliation use dedicated Web Workers. Indexed exact-key matching avoids a quadratic row-by-row comparison path. Results remain in browser memory, processing can be cancelled, and short visible pauses remain possible at the measured maximum.
XLSX generation and download
XLSX generation runs in a dedicated export Web Worker, but the result graph must cross a worker boundary and the complete workbook is assembled in memory before a Blob download begins. At large sizes, this costs substantially more time and memory than matching.
Documented reference run
Measured 200k scenario
These observations come from one run on an Apple M4 Mac with 16 GiB of memory. Each source was a narrow 2.6 MiB CSV with 200,000 exact-match rows and two columns.
Isolated Node measurement
- Parsing + reconciliation
- approximately 0.98 s
- RSS after reconciliation
- approximately 675 MiB
- XLSX generation
- approximately 15.6 s
- Peak RSS after export
- approximately 2.8 GiB
- Generated workbook
- approximately 67.9 MiB
Playwright browser observations
| Stage | Chromium | Firefox | WebKit |
|---|---|---|---|
| Clean run | 3.27 s | 4.54 s | 5.02 s |
| Export/download | 11.61 s | 14.67 s | 22.75 s |
| Longest heartbeat gap | 1.04 s | 1.51 s | 2.14 s |
- Clean run
- Chromium 3.27 sFirefox 4.54 sWebKit 5.02 s
- Export/download
- Chromium 11.61 sFirefox 14.67 sWebKit 22.75 s
- Longest heartbeat gap
- 1.04 s1.51 s2.14 s
What can increase memory and time
- Wider rows and more columns
- Larger raw cell values
- XLSX parsing
- Complex normalizations and comparison rules
- Duplicate or ambiguous key groups
- Large result sets and workbook exports
- Other memory-intensive browser tabs
Current browser evidence
Automated engine coverage and native browser checks are reported separately.
- Chromium
- Playwright Chromium engine
- Representative automated coverage; not an exact native Chrome or Edge test.
- Firefox
- Playwright Firefox 153
- Automated engine coverage; the native Firefox application was not installed.
- WebKit
- Playwright WebKit 26.5
- Automated engine coverage, including the maximum-size scenario.
- Safari 26.5.2
- Native Safari smoke test
- Focused demo workflow only; not the full automated 200k scenario.
Native Chrome, Edge, and Firefox were not tested on the measurement Mac. Playwright WebKit is not presented as an exact native Safari test.
Functional boundaries in V0.5
- CSV and standard XLSX only
- One selected sheet per source
- Deterministic 1:1 matching
- One to five mapped key fields
- Absolute numeric tolerance
- At most one tolerant date-key field
- No percentage tolerance
- No fuzzy or AI matching
- No 1:N, N:1, or N:N matching
- Password-protected workbooks are unsupported
- No saved workflows in V0.5
If a large export cannot finish
Close other memory-intensive tabs. If available, return to the source step and use a smaller source extract. Reconciliation results may remain usable even when a very large XLSX export exceeds browser resources.
Do not interpret a costly export as evidence that matching itself failed; the two stages have different measured costs.
Use the limits to choose the right starting point.
Try the synthetic demo first, or start locally with two CSV or standard XLSX exports.
Read the help guide