Safe Haskell	Safe-Inferred
Language	Haskell98

Text.CSV.Lazy.String

Contents

CSV types
CSV parsing
Pretty-printing
Conversion between standard and simple representations
Selection, validation, and algebra of CSV tables

Description

The CSV (comma-separated value) format is defined by RFC 4180, "Common Format and MIME Type for Comma-Separated Values (CSV) Files", http://www.rfc-editor.org/rfc/rfc4180.txt

This lazy parser can report all CSV formatting errors, whilst also returning all the valid data, so the user can choose whether to continue, to show warnings, or to halt on error.

Valid fields retain information about their original location in the input, so a secondary parser from textual fields to typed values can give intelligent error messages.

In a valid CSV file, all rows must have the same number of columns. This parser will flag a row with the wrong number of columns as a error. (But the error type contains the actual data, so the user can recover it if desired.) Completely blank lines are also treated as errors, and again the user is free either to filter these out or convert them to a row of actual null fields.

Synopsis

CSV types

type CSVTable = [CSVRow] Source

A CSV table is a sequence of rows. All rows have the same number of fields.

type CSVRow = [CSVField] Source

A CSV row is just a sequence of fields.

data CSVField Source

A CSV field's content is stored with its logical row and column number, as well as its textual extent. This information is necessary if you want to generate good error messages in a secondary parsing stage, should you choose to convert the textual fields to typed data values.

Constructors

CSVField
Fields csvRowNum :: !Int csvColNum :: !Int csvTextStart :: !(Int, Int) csvTextEnd :: !(Int, Int) csvFieldContent :: !String csvFieldQuoted :: !Bool
CSVFieldError
Fields csvRowNum :: !Int csvColNum :: !Int csvTextStart :: !(Int, Int) csvTextEnd :: !(Int, Int) csvFieldError :: !String

Instances

Eq CSVField
Show CSVField

CSV parsing

data CSVError Source

A structured error type for CSV formatting mistakes.

Constructors

IncorrectRow
Fields csvRow :: !Int csvColsExpected :: !Int csvColsActual :: !Int csvFields :: [CSVField]
BlankLine
Fields csvRow :: !Int csvColsExpected :: !Int csvColsActual :: !Int csvField :: CSVField
FieldError
Fields csvField :: CSVField
DuplicateHeader
Fields csvColsExpected :: !Int csvHeaderSerial :: !Int csvDuplicate :: !String
NoData

Instances

Eq CSVError
Show CSVError

type CSVResult = [Either [CSVError] CSVRow] Source

The result of parsing a CSV input is a mixed collection of errors and valid rows. This way of representing things is crucial to the ability to parse lazily whilst still catching format errors.

csvErrors :: CSVResult -> [CSVError] Source

Extract just the errors from a CSV parse.

csvTable :: CSVResult -> CSVTable Source

Extract just the valid portions of a CSV parse.

csvTableFull :: CSVResult -> CSVTable Source

Extract the full table, including invalid rows, repaired with padding. and de-duplicated headers.

csvTableHeader :: CSVResult -> [String] Source

The header row of the CSV table, assuming it is non-empty.

parseCSV :: String -> CSVResult Source

A first-stage parser for CSV (comma-separated values) data. The individual fields remain as text, but errors in CSV formatting are reported. Errors (containing unrecognisable rows/fields) are interspersed with the valid rows/fields.

parseDSV :: Bool -> Char -> String -> CSVResult Source

Sometimes CSV is not comma-separated, but delimiter-separated values (DSV). The choice of delimiter is arbitrary, but semi-colon is common in locales where comma is used as a decimal point, and tab is also common. The Boolean argument is whether newlines should be accepted within quoted fields. The CSV RFC says newlines can occur in quotes, but other DSV formats might say otherwise. You can often get better error messages if newlines are disallowed.

Pretty-printing

ppCSVError :: CSVError -> String Source

Some pretty-printing for structured CSV errors.

ppCSVField :: CSVField -> String Source

Pretty-printing for CSV fields, shows positional information in addition to the textual content.

ppCSVTable :: CSVTable -> String Source

Turn a full CSV table back into text, as much like the original input as possible, e.g. preserving quoted/unquoted format of fields.

ppDSVTable :: Bool -> Char -> CSVTable -> String Source

Turn a full CSV table back into text, using the given delimiter character. Quoted/unquoted formatting of the original is preserved. The Boolean argument is to repair fields containing newlines, by replacing the nl with a space.

Conversion between standard and simple representations

fromCSVTable :: CSVTable -> [[String]] Source

Convert a CSV table to a simpler representation, by dropping all the original location information.

toCSVTable :: [[String]] -> ([CSVError], CSVTable) Source

Convert a simple list of lists into a CSVTable by the addition of logical locations. (Textual locations are not so useful.) Rows of varying lengths generate errors. Fields that need quotation marks are automatically marked as such.

Selection, validation, and algebra of CSV tables

selectFields :: [String] -> CSVTable -> Either [String] CSVTable Source

Select and/or re-arrange columns from a CSV table, based on names in the header row of the table. The original header row is re-arranged too. The result is either a list of column names that were not present, or the (possibly re-arranged) sub-table.

expectFields :: [String] -> CSVTable -> Either [String] CSVTable Source

Validate that the named columns of a table have exactly the names and ordering given in the argument.

mkEmptyColumn :: String -> CSVTable Source

A generator for a new CSV column, of arbitrary length. The result can be joined to an existing table if desired.

joinCSV :: CSVTable -> CSVTable -> CSVTable Source

A join operator, adds the columns of two tables together. Precondition: the tables have the same number of rows.

mkCSVField :: Int -> Int -> String -> CSVField Source

Generate a fresh field with the given textual content. The quoting flag is set automatically based on the text. Textual extents are not particularly useful, since there was no original input to refer to.