Short answer
Open big files from disk rather than pasting them. Let Blobster parse up to its raw-first limit (25 MB by default, adjustable up to 250 MB), read anything larger in Raw until you need the tree, and export line-delimited data as NDJSON so it can stream.
Key points
- Files above the raw-first limit open as text first, so the tab stays usable while you decide whether to parse.
- CSV, TSV, NDJSON and JSONL files over about 1 MB stream into the table, and you can stop the import and keep the rows already loaded.
- Tree and table views draw only the rows on screen, and large arrays show 200 entries at a time.
- A single JSON document is parsed in one piece, so the memory your browser gives the tab is the real ceiling.
What Blobster does with a big file
Blobster picks a strategy from the size and format of the file when you open it with the file picker or drag it into the window. All of it runs in your browser. The file is not uploaded, and there is no share link that could publish it.
- JSON, YAML, TOML, XML and other single documents up to the raw-first limit are parsed in a background worker. A progress strip shows the parse and lets you cancel it.
- Files over the limit open in Raw first, with a notice and a
Parse nowbutton. You can read and search the text and build the tree only when you need it. - CSV, TSV, NDJSON and JSONL files larger than about 1 MB stream into the table. Rows appear while the file loads, and
Stop importkeeps what has loaded so far.
Set the limit for your machine
The raw-first limit is under Settings, Advanced, Large files. It accepts 1 to 250 MB and defaults to 25 MB. Raise it on a machine with memory to spare and big files open as a tree without the extra click. Lower it if large files make the tab slow.
Pasting works for big text too, but streaming only applies to files opened from disk. For anything over a few megabytes, open the file instead of copying it through the clipboard.
Read a large tree without expanding everything
Expanding a 50,000-item array is rarely what you want. Blobster shows the first 200 children of a large array or object, with a control that reveals the next 200, and it only draws the rows that are on screen.
Search is usually faster than browsing. Ctrl+F (Cmd+F on macOS) searches keys and values, Enter jumps to the next match, and the tree can be filtered down to the matching branches. On large documents Blobster builds its search index in a worker where it can.
Once you find the part that matters, copy the selected value, subtree or path instead of the whole file.
Where the limits are
A JSON document is one value, so it cannot stream the way NDJSON can. Blobster either parses all of it or shows it as text. How far that goes depends on your machine and on the shape of the data, so there is no single maximum file size.
The browser’s JSON parser stores numbers as 64-bit floating point values. Integers beyond 2^53, such as some database IDs, lose precision in the tree. Raw view keeps the original digits.
For files in the gigabytes, or for processing you want to repeat, use a command-line tool. jq’s --stream option parses input incrementally, and the jq manual describes it as useful for very large inputs.
If you control the export, prefer NDJSON
One JSON array of a million records has to be parsed in one go. The same records written one per line as NDJSON can stream, and a single malformed line does not stop the rest of the file from loading. Blobster shows NDJSON records as table rows and lists bad lines separately, with their line numbers.