Convert structured data with explicit types

Choose the representation

Seven explicit modes are available: CSV/TSV to JSON, JSON to CSV/TSV, YAML 1.2 core to JSON, JSON to YAML, XML to an ordered namespace JSON tree, that exact tree to XML, and XML well-formedness checking with normalized XML output. XML Schema validation and XLSX conversion are not provided by this bounded profile.

Review types and losses before conversion

CSV has no intrinsic types or null value. With no explicit mappings, all columns remain strings. Adding mappings selects only the listed columns: map every column to retain it. Choose delimiter, quote, header and row/object shape, then explicitly map columns to string, exact number, boolean or date. Null tokens, decimal/grouping marks and date profiles are explicit. Significant strings, empty strings, leading zeroes and nulls must not be guessed from their appearance. Preview the selected records, types and loss report, then confirm the exact source and mapping before conversion.

JSON numeric tokens are handled without binary64 rounding by the admitted profile. YAML comments, anchor identity and lexical spelling are not preserved in JSON; aliases expand within fixed bounds. Ordered XML namespace trees retain semantic node order, names and attributes, but parser normalization, entity spelling, quote choices and original encoding are not byte-for-byte round trips. DTDs and external references are disabled.

CSV output is an explicit choice

The spreadsheet-prefix option adds apostrophes to matching formula-looking fields and reports the changed values. Lossless-data output adds no such prefixes and must be imported as data by the receiving application. Quoting and this finite prefix rule do not establish universal spreadsheet safety. Retain the options-only recipe for a typed round trip and inspect the complete downloaded result.

Bounds and local lifetime

Input is limited to 5 MiB, complete output to 25 MiB, recipes to 256 KiB, and table work to 100,000 records, 1,000 columns and 250,000 cells, with depth 128. Preview displays at most 20 records and 25 columns; displayed values and structured text are visibly shortened. The confirmed download contains the complete admitted output, never a preview prefix.

Each explicit action has one original twenty-second window. Local reads use at most five seconds and producer plus fresh source verification at most fifteen seconds within that same endpoint. Once complete preparation and fresh verification finish within that endpoint, accepted data remains available for review until an edit, Cancel, Clear or workflow retirement. Each explicitly created temporary download link expires after twenty seconds; a later explicit download can create another link for the same accepted data. Declared UTF-8, UTF-16LE/BE, Shift-JIS, EUC-JP, ISO-2022-JP and EUC-KR input is admitted only when the browser passes the selected decoding sentinels; there is no fallback substitution. Output is UTF-8.

Edits, Cancel, Clear and changing workflow withdraw the old result before cleanup. Replacement waits for known original read, worker and download cleanup; uncertainty blocks new work until reload. Cancellation can wait for a running parser to finish. Logical closure does not prove native drain, physical memory erasure, device capacity or complete format conformance. An already downloaded file remains on your device.

Fixed complete downloads

Results use webut-structured-result.json, .csv, .tsv, .yaml or .xml. The explicit recipe is webut-structured-recipe.json and the private preview/loss report is webut-structured-report.json. Recipes and reports can contain private column names, tokens or preview values; they are downloaded only when explicitly chosen. No source filename is used.

Broader XLSX, XML Schema/local reference packs, split-table ZIP, legacy output encoders, regional, native-language and receiving-application acceptance remain required. These bounded operations do not complete those broader capabilities or the WebUT product.