Data Importer & Cleaner is a Windows desktop application for turning inconsistent CSV and Excel files into clean, validated and reusable business data through a safe, reviewable workflow.

Important

Version 1.0.1 is the current stable release. M1–M8 and the audit-hardening pass are complete, and the self-contained Windows package is available from GitHub Releases.

Business data often arrives with inconsistent headers, whitespace, date formats, decimal separators, country names, missing values and duplicates. Fixing these files manually is slow and difficult to audit. This project provides a deterministic desktop workflow that preserves the source data and makes every transformation reviewable before export.

The repository is also a portfolio case study in layered .NET desktop development, data processing, validation, persistence, automated testing and release engineering.

Validation and cleaning remain separate. The validation engine can run before and after cleaning so that users can distinguish source problems from issues that remain in the final dataset.

The profiling view combines inferred column statistics, validation settings and explicit source-to-target mappings in one reviewable step.

Every deterministic transformation is recorded with its row, column, rule, source value, resulting value and explanation.

The final summary reports valid, invalid and modified rows and keeps a local record of the generated output without persisting imported row contents.

Area Status Target milestone Layered solution and WPF shell Complete M1 Dependency injection and logging Complete M1 SQLite model and initial migration Complete M1 Domain foundations and processing contracts Complete M1 CSV import and preview Complete M2 XLSX import and worksheet selection Complete M2 Data profiling and column mapping Complete M3 Validation engine Complete M4 Cleaning engine and before/after review Complete M5 Deterministic duplicate detection Complete M6 CSV/XLSX/SQLite export and history Complete M7 Portfolio release Complete M8

See the roadmap and project plan for scope and delivery details.

Immutable source: the input file is never modified or overwritten.

Traceable values: every cell preserves its source, parsed and current value.

Deterministic behavior: transformations and duplicate matching are predictable and reproducible.

Explicit culture: ambiguous numbers and dates are never silently interpreted without culture settings.

Review before export: users can inspect original and transformed values before producing output.

Privacy by default: imported row contents are not persisted or logged by default.

Focused MVP: AI cleaning, fuzzy matching, cloud sync and multi-user features remain outside v1.0.

DataCleaner.Domain contains framework-independent business concepts.

DataCleaner.Application declares workflows and ports for files and persistence.

DataCleaner.Infrastructure contains EF Core, SQLite and CSV/XLSX adapters.

DataCleaner.App is the WPF composition and presentation layer.

The domain avoids names such as DataSet and DataRow that collide with System.Data. Technical data types and semantic types are modeled separately. More detail is available in Architecture and Domain model.

.NET 10 and C#

WPF with an MVVM-oriented presentation layer

Microsoft.Extensions.Hosting, dependency injection and logging

Entity Framework Core and SQLite

Open XML SDK for XLSX workbooks

xUnit and coverlet

GitHub Actions and Dependabot

The self-contained win-x64 package does not require a separately installed .NET runtime. The v1.0.1 release verification produced a zero-warning Release build, 63 passing tests, 27 passing documentation checks and enforced per-assembly coverage baselines. The archive includes samples/synthetic-customers.csv, enabling a download → unzip → run → try demo workflow without returning to the repository. The complete synthetic 50,000-row pipeline finished in 2.853 seconds on the documented development machine; see performance evidence. Verification details are recorded in the v1.0.1 release notes.

Download DataImporterCleaner-v1.0.1-win-x64.zip together with its published SHA-256 checksum.

Use synthetic-customers.csv for a safe first run. It intentionally contains whitespace, country aliases, invalid values and duplicates. Generate larger deterministic samples with scripts/Generate-SyntheticDataset.ps1.

Windows 10 or Windows 11

PowerShell 7 or Windows PowerShell

Git, if you intend to contribute

The required SDK feature band is pinned in global.json.

The application creates its local SQLite database under the current user's local application-data directory.

Every pull request is expected to meet these checks:

restore succeeds;

Release build produces zero warnings and zero errors;

all automated tests pass;

no real customer or production data is committed;

user-visible or architectural changes update the relevant documentation;

the changelog is updated when the change is release-relevant.

Start with the documentation index. Key documents include:

Bug reports, scoped feature proposals and documentation corrections are welcome through the supplied GitHub issue forms. Because the source remains all-rights-reserved, code or documentation pull requests require prior written permission; see CONTRIBUTING.md.

For help choosing the correct channel, see SUPPORT.md. Please report security vulnerabilities privately as described in SECURITY.md.

Copyright is all rights reserved. The source is available for inspection but is not offered under an open-source license. See LICENSE.