Convert CSV to JSON with Header Keys, Quoted Fields and Malformed Rows

In a practical coding round, write a converter from CSV to JSON. The input is CSV text whose first line is a header naming the columns. Every following record becomes one JSON object that maps each column name to that record's value, and the output is a JSON array of those objects. The report names only the task, so agree on the parsing and output rules before coding: CSV has edge cases that break a naive implementation. ```python def csv_to_json(csv_text: str) -> str: ... ``` Illustrative input: ```text id,name,city 1,Ana,"Austin, TX" 2,"Li ""Leo"" Chen", ``` Expected output: ```json [ {"id": "1", "name": "Ana", "city": "Austin, TX"}, {"id": "2", "name": "Li \"Leo\" Chen", "city": ""} ] ``` ```hint A comma is not always a separator Decide how the parser knows whether it is inside a quoted field, and what that means for the commas, quote characters and line breaks it meets there. ``` ```hint Escaping goes both ways Values can contain characters that are special in JSON as well as in CSV. Decide which layer is responsible for escaping them in the output. ``` ### Clarifying Questions - Should values stay strings, or should numbers, booleans and empty fields be converted (to numbers, `true`/`false`, `null`)? - What should happen when a row has more or fewer fields than the header, or when header names repeat or are empty? - Can quoted fields contain commas, doubled quotes and line breaks? Are line endings `\n`, `\r\n` or both, and can the file start with a byte-order mark? - Is the input small enough to hold in memory, or must the converter stream a large file? - Should each object keep the header's column order, and should the output be one JSON array or one object per line? ### What a Strong Answer Covers - Settles typing, malformed-row and output-shape rules before coding - A quote-aware parser that handles commas, doubled quotes and line breaks inside quoted fields, and both line-ending styles - Correct JSON produced by a serializer, with key order preserved and proper escaping - Clear errors for malformed input (an unterminated quote, a row with the wrong number of fields) that include the line number - Tests for the edge cases, and a plan for streaming large files ### Follow-up Questions - The file is too large to fit in memory. How do you convert it? - Infer column types (integers, decimals, booleans, nulls) consistently per column rather than per cell. What rules do you use, and which values must stay strings? - Header names such as `address.city` should become nested objects. How do you build them, and what conflicts can occur? - Would you use your own parser or a standard library parser in production, and what would you test before trusting either?