Convert CSV, JSON, and spreadsheet data in the browser

The active structured-data inputs are JSON, CSV, TSV, and TXT, and each row in the queue exposes the active destinations among those same four. That covers the overwhelming majority of real data-shuffling: an API response that needs to open in a spreadsheet, an export that needs a different delimiter, a log that needs to become rows.

Spreadsheet workbooks themselves — XLSX and its relatives — are reference guides in the format directory rather than active converters in this release. The workflow is still simple: use File > Download or Save As in Excel, Google Sheets, or LibreOffice to export the sheet as CSV, then bring that CSV to Novus Convert for whatever transformation comes next. One sheet per CSV is the rule, since the format has no concept of tabs.

Drop the data files into the queue at convert.novusstreamsolutions.com/convert. As with every route, the app inspects the actual content rather than believing the extension — useful with data files in particular, because they are constantly mislabeled. An export named .csv that is really tab-separated, or a .json file that is actually a JavaScript snippet with trailing commas, gets flagged instead of silently producing garbage.

Structured text accepts files up to 25 MB, which is roughly hundreds of thousands of rows for a typical export. If a file is over the line, trim it at the source — filter the export to the date range or columns you need — rather than hand-splitting a giant file and hoping the pieces stay consistent.

CSV and TSV are flat grids: rows and columns, nothing else. JSON is a tree that can nest objects inside objects to any depth. Conversion between them is only clean when the JSON is already table-shaped — an array of objects where every object has the same flat set of keys. That shape maps perfectly: each object becomes a row, each key becomes a column header.

Deeply nested JSON — a customer object holding an array of orders, each holding an array of items — has no faithful flat representation, and no converter can invent one without making decisions for you. If your JSON nests, restructure it to a flat array of records before converting, or accept that the tabular output will represent nested values in a flattened, less convenient form. Going the other direction is safer by nature: any CSV or TSV can become a JSON array of flat objects without losing anything.

The comma in CSV is a landmine when the data itself contains commas — addresses, product descriptions, names written surname-first. Well-formed CSV wraps such values in quotes, and the parser honors quoting and embedded line breaks, but plenty of tools downstream get quoting subtly wrong. When your text is comma-heavy, converting to TSV sidesteps the whole class of problem, because literal tab characters almost never appear inside real-world values.

This is also the fix for regional spreadsheet quirks. Some locales expect semicolon-separated files and will happily dump every value of a comma-separated file into column A. If a colleague reports that your CSV "opens as one column," send them a TSV instead — spreadsheet apps detect tabs reliably across locales — or have them use their app's explicit import dialog rather than double-clicking the file.

Encoding problems are the ones that hide longest. A file converts, validates, opens — and then every accented name reads as é and every curly quote as a triplet of junk characters. The output Novus Convert produces is UTF-8, today's universal default. Trouble almost always enters upstream, when a legacy tool exported the source in a regional encoding, or downstream, when an app opens UTF-8 as something else.

The classic downstream case is Excel double-clicking a UTF-8 CSV and mangling non-ASCII characters. The cure is Excel's import flow (Data > From Text/CSV), which lets you state the encoding explicitly. Upstream, if the source file already shows mangled characters before you convert it, fix the export first — a converter faithfully preserves bytes, including faithfully preserving damage.

A data row only offers its download after the output passes a parse check, so a malformed source — truncated JSON, a CSV with a broken quote sequence — fails at the gate instead of producing a file that explodes in some importer next week. Treat a validation failure on a data file as information about the source, and inspect it in a text editor before blaming the route.

Then do the check no automation can: open the output where it will actually live. Load the CSV into the spreadsheet and confirm the column count, the header row, and a few values deep in the file, not just the top. Feed the JSON to the script or endpoint that will consume it. For high-stakes data, close the loop with a round trip — convert the output back toward the source format and confirm row counts and spot-checked values survived both directions. Structured text carries a generous allowance of 50 downloads per day, so verification passes cost you nothing.