How to convert CSV to JSON on iPhone

Take this CSV:

name,city,age
Ada,London,36
Grace,New York,45

You get this JSON — an array, one object per data row, with the header row as the keys:

[
  {
    "age" : "36",
    "city" : "London",
    "name" : "Ada"
  },
  {
    "age" : "45",
    "city" : "New York",
    "name" : "Grace"
  }
]

It is pretty-printed, so it is readable as it stands, and the keys come out in alphabetical order rather than column order. JSON objects have no inherent order, so nothing has been lost — but if you were expecting the columns in their original sequence, that is why they are not.

Notice "age" : "36" — quoted. Not 36. Convexy does not guess types, and it is worth understanding why, because the converters that do guess will eventually ruin your data.

A CSV file contains no type information whatsoever. Every cell is text. When a converter decides that 36 should become a number, it is applying a heuristic — and the heuristic is wrong in exactly the cases that matter:

• Leading zeros vanish. A US zip code 02134 becomes the number 2134. A product SKU 00417 becomes 417. The data is now wrong and nothing announces it.
• Phone numbers and IDs get mangled. A long account number can exceed the range where a double-precision number is exact, and quietly changes value.
• Version strings and dates like 1.10 become 1.1.
• Blank cells become an argument: empty string, zero, or null?

Convexy takes the position that a converter should not silently rewrite your data. You get exactly what was in the file, as text. If your consuming code needs numbers, it can cast them — that is a decision your code is qualified to make and a file converter is not.

CSV looks trivial and is not. There is no single standard; it is a folk format. These are the things that will bite you, and how Convexy behaves in each case:

• Commas inside quoted fields. "Smith, John",42 is two fields, not three. This is handled: text inside quotes is not split on commas.
• Newlines inside quoted fields. A cell containing a line break — an address, a comment field exported from a survey tool — will break the row. The file is read line by line, so a field spanning two lines is read as two rows, and everything after it is misaligned. This is the single most common way a real-world CSV export defeats a simple converter, and Convexy is a simple converter here. If your data has multi-line cells, check the output before trusting it.
• Literal quote characters. Double quotes are treated as field delimiters, so a quote character inside a value does not survive into the JSON. Values with quoted speech in them will come out with the quotation marks removed.
• No header row. The first line is always treated as the header. If your file starts straight into data, that first record becomes your keys, and you lose it as data. Add a header row first.
• Rows with too few values. Missing trailing columns are filled with empty strings, so the object always has every key.
• Rows with too many values. Anything beyond the number of headers is dropped.
• Duplicate column names. Two columns with the same header collapse into one key — the later one wins. Rename the columns first.
• Semicolon-separated files. Exports from European locales often use ; as the separator. Convexy splits on commas only, so such a file converts to a single column of long strings. Re-export it comma-separated.
• Whitespace around values is trimmed.

The failure mode with tabular data is not a crash — it is a file that looks fine and is subtly misaligned. Open the JSON and confirm the record count matches your row count, and that the last record is complete. If those two things are right, the parse held. The reverse conversion is available too, and round-tripping is a decent sanity check.