Dataset & CSV Cleaner → Excel — Filter, Dedupe, Japanese CSV

Turn messy data into a clean Excel file in one run. Works on any Apify dataset (e.g. a scraper's results), CSV / TSV / JSON / JSON Lines / Excel files, or JSON you paste in. $0.05 per 1,000 output rows.

• Flatten nested JSON into tidy columns (address.city), lists into one cell or one column per item
• Filter with 15 operators — numbers like 1,200 and dates like 2026/10/07 or 2026年10月7日 are compared correctly
• Remove duplicates by one or more fields (or the whole row)
• Sort naturally: 2 < 10 < 1,000, item2 < item10, Japanese text in Japanese order, empty values last
• Choose, order and rename columns (Japanese column names are fine)
• Clean text: trim spaces (incl. full-width 　), collapse spaces, normalize full-width / half-width characters (ＡＢＣ１２３ → ABC123, ｶﾀｶﾅ → カタカナ) so that Japanese data matches and dedupes correctly
• Excel output (.xlsx) with a frozen header row, auto filter and fitted column widths; "1,200" becomes a real number, while zip codes and phone numbers with leading zeros stay text
• CSV that opens in Excel without garbled Japanese (UTF-8 with BOM), and Shift_JIS CSV for older Japanese systems
• Reads Shift_JIS / EUC-JP / GBK / Big5 / EUC-KR CSV files (auto-detected)

Why?

Exported CSV files often look broken when opened in Excel: Japanese, Chinese and Korean text turns into unreadable characters, nested fields appear as [object Object], and numbers like 1,200 are text. This Actor fixes all of that and adds the usual clean-up steps — filtering, deduplication, sorting and column selection — so the result can be shared directly.

Typical uses

• Make a scraper's dataset Excel-ready: flatten, keep 5 columns, rename them, dedupe, sort
• Merge a dataset with CSV/Excel files and remove duplicates
• Convert a Shift_JIS CSV from a Japanese system to UTF-8 / Excel, or the other way round
• Filter a large export to the rows you need (e.g. price >= 1000 and prefecture in 東京都, 大阪府)
• Clean data before loading it into an AI / RAG pipeline (JSON Lines output)

Input

Field	Description
datasetId	An Apify dataset (read-only access to that dataset only)
fileUrls	CSV, TSV, JSON, JSON Lines or .xlsx file URLs
data	Rows as a JSON array
flattenNested, arrayMode	Nested fields → columns
trimSpaces, normalizeWidth, collapseSpaces, emptyToNull	Text clean-up
filters, filterMode, caseSensitive	Keep matching rows
dedupeBy, sortBy, offset, limit	Dedupe, sort, slice
fields, renameFields	Columns to keep (in order) and new names
outputFormats	xlsx, csv-excel, csv-sjis, csv, tsv, json, jsonl

Filter example (rows from Tokyo or Osaka with a price of at least 1,000):

{
    "filters": [
        { "field": "address.pref", "operator": "in", "value": "東京都, 大阪府" },
        { "field": "price", "operator": "gte", "value": "1000" }
    ],
    "filterMode": "all"
}

Operators: equals, notEquals, contains, notContains, startsWith, endsWith, in, regex, gt, gte, lt, lte, isEmpty, notEmpty.

Processing order: load → flatten → clean text → filter → dedupe → sort → offset/limit → select & rename columns.

Output

• Files in the key-value store (Output tab → "Output files"): data.xlsx, data-excel.csv, … with download links
• Cleaned rows in the dataset (optional), so Apify's own exports and integrations work too
• SUMMARY: rows in, rows after filters, duplicates removed, rows out, columns, files

Example: the prefilled rows with normalizeWidth on, dedupe by name, a date filter (updated ≥ 2026/10/01), sorting by -price and two renamed columns. ＡＢＣ商店 and ABC商店 are recognised as the same shop and kept once (CSV for Excel, first lines):

name,価格,tags,都道府県,address.city,updated
テスト(株) 😀,15000,,北海道,札幌市,2026-10-01
ABC商店,"1,200",食品; セール,東京都,千代田区,2026年10月7日

Pricing

Event	Price
Output row	$0.00005 ($0.05 per 1,000 rows)
Run start	$0.002

You pay for the rows in the result, not for rows that were filtered out or removed as duplicates. All output formats are included.

Notes

• Only the dataset and files you provide are read.
• Characters that do not exist in Shift_JIS (such as emoji) are written as ? in the Shift_JIS CSV; the run log tells you how many. The Excel file and the UTF-8 CSV keep everything.
• Large inputs are held in memory: about 1 GB per ~200,000 typical rows. Increase the memory for bigger datasets.