How the canonical tree prevents conversion sprawl
The tool does not maintain a different converter for every pair. JSON, YAML, and Properties are parsed into ordered mapping, sequence, and scalar nodes. Structured JSON and YAML retain that hierarchy; their flattened shapes, Properties, dotenv, and shell use dot paths and bracket indexes. This design gives all fifteen Phase 1 source-to-target combinations the same limits, collision checks, type policy, and deterministic ordering.
The document root must be a mapping. Top-level arrays and scalars are rejected because deployment configuration normally needs named fields. Objects preserve source key order, arrays preserve item order, and prototype-related names stay inert data rather than being merged into JavaScript objects used by application logic.
Flattening rules are intentionally strict
A nested key such as database.host becomes a dot path. An array item such as the first service name becomes services[0].name. Numeric dot segments are object keys, not inferred arrays. Sparse indexes, empty path segments, invalid brackets, and a scalar/container prefix collision are blocking errors because guessing could silently deploy the wrong value.
Nested and literal dotted keys can coexist. Flattening retains literal punctuation and blocks duplicate output keys, with a warning that the original key boundaries may not be recoverable. Input keys lets you preserve source keys or explicitly expand dotted YAML/JSON keys into mappings. Environment targets additionally normalize boundaries and camel case to uppercase underscores. If two source paths produce the same environment name, conversion stops and identifies the paths—not their values.
Types are preserved only where the target can carry them
Both shapes of JSON and YAML retain strings, numbers, booleans, and null; the shape choice changes only the key layout. Properties values are strings, so the default Properties-to-JSON or YAML conversion remains string-safe. Optional inference recognizes only exact lowercase booleans, exact null, and strict JSON numbers. Identifiers such as 00123, dates, empty values, padded values, NaN, and Infinity remain strings.
Properties, dotenv, and shell cannot preserve native types. Typed JSON or YAML values become strings, and null becomes the literal text null, each with a warning. Empty objects and arrays have no safe representation in flat targets; choose structured JSON or YAML to retain them. The round-trip assessment is conservative: a warning means you should not expect an equivalent typed tree without review.
Generic value handling
Preserve keeps numbers, booleans, null, and strings typed when the target supports them. Strings explicitly converts every scalar to text and reports any changed values before output.
Properties, dotenv, and shell are inherently string-oriented. JSON and YAML can retain native scalar types, so use the explicit value policy that matches the consuming application.
Local execution, limits, and safe review
Conversion runs in a cancellable same-origin Web Worker so large valid files do not freeze the interface. The worker enforces a 1 MiB input limit, 64 nesting levels, 50,000 canonical nodes, 25,000 flattened entries, and 5 MiB generated output. YAML accepts one core-schema document, rejects duplicate keys, custom tags, and merge keys, and bounds alias expansion.
No configuration is sent to an application server or stored in cookies, Web Storage, IndexedDB, URLs, logs, analytics, advertising, or the offline cache. Clear terminates the active worker and removes source, output, diagnostics, filename metadata, and download URLs. JavaScript strings are garbage-collected, so browser software cannot promise deterministic physical memory zeroization. Always inspect generated output and keep secrets in your approved secret-management system.
