CSVKit for Objective-C

CSVKit is a comma-separated value (CSV) parser for Objective-C.

Copyright © 2011 Booyah, Inc.

Jon Parise <[email protected]>

Overview

CSVKit provides block-based CSV parsers for Objective-C. The internal parsing routines are written in C for speed (and are largely adapted from Python's csv module). The parser calls user-supplied blocks as it incrementally parses the source data.

Because there's no true "standard" for CSV-style data, the parsing rules can be configured via "dialects". Dialects define things like the format's field delimiter and quoting rules.

Currently, values are always either NSString or NSNumber objects. Strings are always treated as UTF8-encoded, and numbers are always interpreted as doubles. Numeric values are only recognized if the dialect provides a way to differentiate numbers from strings (e.g. CSVQuoteStyleNonNumeric).

Installation

Simply include these source files in your project:

• CSVKit.h
• CSVKit.m

The repository also includes a SenTestingKit-compatible unit test:

• CSVKitTests.h
• CSVKitTests.m

Usage

Dialects

Dialects dictate the parsing rules and are defined by the CSVDialect struct. You can define your own rules by fillout out a new CSVDialect object, or you can use one of the default dialects provided by CSVKit.

• CSVExcelDialect - Excel-generated CSV data
• CSVExcelTabDialect - Excel-generated TAB-delimited data

Field-Based Parsing

The lowest level of parsing occurs at the field level. Field parsing is available through the -parseFieldsFrom* family of methods on the CSVParser class.

Block Callback

void (^)(id value, NSUInteger index, BOOL *stop)

• value - Object representing the parsed field's value
• index - 0-based field index; NSUIntegerMax at end of line
• stop - Flag that will stop processing if set to YES

It's important to handle the index == NSUIntegerMax case as that's the only way to determine when the parser has completed parsing the current line's values. Also note that value will be nil in this case.

Row-Based Parsing

It's often more convenient to work with complete rows of values. Row parsing is available via the -parseRowsFrom* methods.

Block Callback

void (^)(NSArray *row, BOOL *stop)

• row - Array of parsed values
• stop - Flag that will stop processing if set to YES

Object-Based Parsing

Lastly, it's possible to parse rows of values directly into object properties. This is a fast and convenient way to build data or model objects directly from CSV source data.

Because this style of parsing is a bit more specialized, you need to use the CSVObjectParser subclass of CSVParser. This class allows you to specify both an object class (to use when constructing new per-row object instances) as well as an array of property names.

- (id)initWithDialect:(const CSVDialect *)dialect
          objectClass:(Class)objectClass
        propertyNames:(NSArray *)propertyNames;

If objectClass is not specified, we default to NSMutableDictionary. If propertyNames is not specified, we use the fields from the first row of data as the property names.

Object instances are allocated using -alloc and initialized using -init:

[[objectClass alloc] init]

Property values are set using -setValue:forKey:. In the future, we may also support key paths (-setValue:forKeyPath:) via an optional flag of some sort, but key path-based assignment is quite a bit slower at runtime, so it is not the default behavior.

Block Callback

void (^)(id object, BOOL *stop)

• object - Parsed object value
• stop - Flag that will stop processing if set to YES

Future Ideas

• Incremental (stream-based) parsing
• Dialect guessing
• Richer numeric types
• Object caching for frequently-occuring values
• Optimized NSDictionary parsing if there's enough interest