Tabular data is routinely transferred on the web in a variety of formats, including variants on CSV, tab-delimited files, fixed field formats, spreadsheets, HTML tables, and SQL dumps. This document outlines a data model, or infoset, for tabular data and metadata about that tabular data that can be used as a basis for validation, display, or creating other formats. It also contains some non-normative guidance for publishing tabular data as CSV and how that maps into the tabular data model.

An annotated model of tabular data can be supplemented by separate metadata about the table. This specification defines how implementations should locate that metadata, given a file containing tabular data. The standard syntax for that metadata is defined in [[!tabular-metadata]]. Note, however, that applications may have other means to create annotated tables, e.g., through some application specific API-s; this model does not depend on the specificities described in [[!tabular-metadata]].

The CSV on the Web Working Group was chartered to produce a Recommendation "Access methods for CSV Metadata" as well as Recommendations for "Metadata vocabulary for CSV data" and "Mapping mechanism to transforming CSV into various Formats (e.g., RDF, JSON, or XML)". This document aims to primarily satisfy the first of those Recommendations (see ), though it also specifies an underlying model for tabular data and is therefore a basis for the other chartered Recommendations.

This definition of CSV used in this document is based on IETF's [[!RFC4180]] which is an Informational RFC. The working group's expectation is that future suggestions to refine RFC 4180 will be relayed to the IETF (e.g. around encoding and line endings) and contribute to its discussions about moving CSV to the Standards track.

Many files containing tabular data embed metadata, for example in lines before the header row of an otherwise standard CSV document. This specification does not define any formats for embedding metadata within CSV files, aside from the titles of columns in the header row which is defined in CSV. We would encourage groups that define tabular data formats to also define a mapping into the annotated tabular data model defined in this document.

Introduction

Tabular data is data that is structured into rows, each of which contains information about some thing. Each row contains the same number of cells (although some of these cells may be empty), which provide values of properties of the thing described by the row. In tabular data, cells within the same column provide values for the same property of the things described by each row. This is what differentiates tabular data from other line-oriented formats.

Tabular data is routinely transferred on the web in a textual format called CSV, but the definition of CSV in practice is very loose. Some people use the term to mean any delimited text file. Others stick more closely to the most standard definition of CSV that there is, [[!RFC4180]]. Appendix A describes the various ways in which CSV is defined. This specification refers to such files, as well as tab-delimited files, fixed field formats, spreadsheets, HTML tables, and SQL dumps as tabular data files.

In , this document defines a model for tabular data that abstracts away from the varying syntaxes that are used for when exchanging tabular data. The model includes annotations, or metadata, about collections of individual tables, rows, columns, and cells. These annotations are typically supplied through separate metadata files; defines how these metadata files can be located, while [[!tabular-metadata]] defines what they contain.

Once an annotated table has been created, it can be processed in various ways, such as display, validation, or conversion into other formats. This processing is described in .

This specification does not normatively define a format for exchanging tabular data. However, it does provide some best practice guidelines for publishing tabular data as CSV, in section , and for parsing both this syntax and those similar to it, in .

This specification makes use of the compact IRI Syntax; please refer to the Compact IRIs from [[!JSON-LD]].

This specification makes use of the following namespaces:

csvw:
http://www.w3.org/ns/csvw#
dc:
http://purl.org/dc/terms/
rdf:
http://www.w3.org/1999/02/22-rdf-syntax-ns#
rdfs:
http://www.w3.org/2000/01/rdf-schema#
schema:
http://schema.org/
xsd:
http://www.w3.org/2001/XMLSchema#

Typographical conventions

Tabular Data Models

This section defines an annotated tabular data model: a model for tables that are annotated with metadata. Annotations provide information about the cells, rows, columns, tables, and groups of tables with which they are associated. The values of these annotations may be lists, structured objects, or atomic values. Core annotations are those that affect the behavior of processors defined in this specification, but other annotations may also be present on any of the components of the model.

Annotations may be described directly in [[!tabular-metadata]], be embedded in a tabular data file, or created during the process of generating an annotated table.

String values within the tabular data model (such as column titles or cell string values) MUST contain only Unicode characters.

In this document, the term annotation refers to any metadata associated with an object in the annotated tabular data model. These are not necessarily web annotations in the sense of [[annotation-model]].

Table groups

A group of tables comprises a set of annotated tables and a set of annotations that relate to that group of tables. The core annotations of a group of tables are:

Groups of tables MAY in addition have any number of annotations which provide information about the group of tables. Annotations on a group of tables may include:

When originating from [[!tabular-metadata]], these annotations arise from common properties defined on table group descriptions within metadata documents.

Tables

An annotated table is a table that is annotated with additional metadata. The core annotations of a table are:

The table MAY in addition have any number of other annotations. Annotations on a table may include:

When originating from [[!tabular-metadata]], these annotations arise from common properties defined on table group descriptions within metadata documents.

Columns

A column represents a vertical arrangement of cells within a table. The core annotations of a column are:

Columns MAY in addition have any number of other annotations, such as a description. When originating from [[!tabular-metadata]], these annotations arise from common properties defined on table group descriptions within metadata documents.

Rows

A row represents a horizontal arrangement of cells within a table. The core annotations of a row are:

Rows MAY have any number of additional annotations. The annotations on a row provide additional metadata about the information held in the row, such as:

Neither this specification nor [[!tabular-metadata]] defines a method to specify such annotations. Implementations MAY define a method for adding annotations to rows by interpreting notes on the table.

Cells

A cell represents a cell at the intersection of a row and a column within a table. The core annotations of a cell are:

There presence or absence of quotes around a value within a CSV file is a syntactic detail that is not reflected in the tabular data model. In other words, there is no distinction in the model between the second value in a,,z and the second value in a,"",z.

Cells MAY have any number of additional annotations. The annotations on a cell provide metadata about the value held in the cell, particularly when this overrides the information provided for the column and row that the cell falls within. Annotations on a cell might be:

Neither this specification nor [[!tabular-metadata]] defines a method to specify such annotations. Implementations MAY define a method for adding annotations to cells by interpreting notes on the table.

Datatypes

Columns and cell values within tables may be annotated with a datatype which indicates the type of the values obtained by parsing the string value of the cell.

Datatypes are based on a subset of those defined in [[!xmlschema11-2]]. The annotated tabular data model limits cell values to have datatypes as shown on the diagram:

Built-in Datatype Hierarchy diagram
Diagram showing the built-in datatypes, based on [[!xmlschema11-2]]; names in parentheses denote aliases to the [[!xmlschema11-2]] terms (see the diagram in SVG or PNG formats)

The core annotations of a datatype are:

If the id of a datatype is that of a built-in datatype, the values of the other core annotations listed above MUST be consistent with the values defined in [[!xmlschema11-2]] or above. For example, if the id is xsd:integer then the base must be xsd:decimal.

Datatypes MAY have any number of additional annotations. The annotations on a datatype provide metadata about the datatype such as title or description. These arise from common properties defined on datatype descriptions within metadata documents, as defined in [[!tabular-metadata]].

The id annotation may reference an XSD, OWL or other datatype definition, which is not used by this specification for validating column values, but may be useful for further processing.

Length Constraints

The length, minimum length and maximum length annotations indicate the exact, minimum and maximum lengths for cell values.

The length of a value is determined as defined in [[!xmlschema11-2]], namely as follows:

  • if the value is null, its length is zero.
  • if the value is a string or one of its subtypes, its length is the number of characters (ie [[!UNICODE]] code points) in the value.
  • if the value is of a binary type, its length is the number of bytes in the binary value.

If the value is a list, the constraint applies to each element of the list.

Value Constraints

The minimum, maximum, minimum exclusive, and maximum exclusive annotations indicate limits on cell values. These apply to numeric, date/time, and duration types.

Validation of cell values against these datatypes is as defined in [[!xmlschema11-2]]. If the value is a list, the constraint applies to each element of the list.

Locating Metadata

As described in , tabular data may have a number of annotations associated with it. Here we describe the different methods that can be used to locate metadata that provides those annotations.

In the methods of locating metadata described here, metadata is provided within a single document. The syntax of such documents is defined in [[!tabular-metadata]]. Metadata is located using a specific orer of precedence:

  1. metadata supplied by the user of the implementation that is processing the tabular data, see .
  2. metadata in a document linked to using a Link header associated with the tabular data file, see .
  3. metadata located through a site-wide location configuration, see .
  4. metadata embedded within the tabular data file itself, see .

Processors MUST use the first metadata found for processing a tabular data file by using overriding metadata, if provided. Otherwise processors MUST attempt to locate the first metadata document from the Link header or the metadata located through site-wide configuration. If no metadata is supplied or found, processors MUST use embedded metadata. If the metadata does not originate from the embedded metadata, validators MUST verify that the table group description within that metadata is compatible with that in the embedded metadata, as defined in [[!tabular-metadata]].

Overriding Metadata

Processors SHOULD provide users with the facility to provide their own metadata for tabular data files that they process. This might be provided:

For example, a processor might be invoked with:

$ csvlint data.csv --datatypes:string,float,string,string
        

to enable the testing of the types of values in the columns of a CSV file, or with:

$ csvlint data.csv --schema:schema.json
        

to supply a schema that describes the contents of the file, against which it can be validated.

Metadata supplied in this way is called overriding, or user-supplied, metadata. Implementations SHOULD define how any options they define are mapped into the vocabulary defined in [[!tabular-metadata]]. If the user selects existing metadata files, implementations MUST NOT use metadata located through the Link header (as described in ) or site-wide location configuration (as described in ).

Users should ensure that any metadata from those locations that they wish to use is explicitly incorporated into the overriding metadata that they use to process tabular data. Processors may provide facilities to make this easier by automatically merging metadata files from different locations, but this specification does not define how such merging is carried out.

Link Header

If the user has not supplied a metadata file as overriding metadata, described in , then when retrieving a tabular data file via HTTP, processors MUST retrieve the metadata file referenced by any Link header with:

so long as this referenced metadata file describes the retrieved tabular data file (ie, contains a table description whose url matches the request URL).

If there is more than one valid metadata file linked to through multiple Link headers, then implementations MUST use the metadata file referenced by the last Link header.

For example, when the response to requesting a tab-separated file looks like:

HTTP/1.1 200 OK
Content-Type: text/tab-separated-values
...
Link: <metadata.json>; rel="describedBy"; type="application/csvm+json"
        

an implementation must use the referenced metadata.json to supply metadata for processing the file.

If the metadata file found at this location does not explicitly include a reference to the requested tabular data file then it MUST be ignored. URLs MUST be normalized as described in .

The Link header of the metadata file MAY include references to the CSV files it describes, using the describes relationship. For example, in the countries' metadata example, the server might return the following headers:

Link: <http://example.org/countries.csv>; rel="describes"; type="text/csv"
Link: <http://example.org/country_slice.csv>; rel="describes"; type="text/csv"

However, locating the metadata SHOULD NOT depend on this mechanism.

Site-wide Location Configuration

If the user has not supplied a metadata file as overriding metadata, described in , and no applicable metadata file has been discovered through a Link header, described in , processors MUST attempt to locate a metadata documents through site-wide configuration.

In this case, processors MUST retrieve the file from the well-known URI /.well-known/csvm. (Well-known URIs are defined by [[!RFC5785]].) If no such file is located (i.e. the response results in a client error 4xx status code or a server error 5xx status code), processors MUST proceed as if this file were found with the content:

{+url}-metadata.json
csv-metadata.json
        

The response to retrieving /.well-known/csvm MAY be cached, subject to cache control directives. This includes caching an unsuccessful response such as a 404 Not Found.

This file MUST contain a URI template, as defined by [[!URI-TEMPLATE]], on each line. Starting with the first such URI template, processors MUST:

  1. Expand the URI template, with the variable url being set to the URL of the requested tabular data file (with any fragment component of that URL removed).
  2. Resolve the resulting URL against the URL of the requested tabular data file.
  3. Attempt to retrieve a metadata document at that URL.
  4. If no metadata document is found at that location, or if the metadata file found at the location does not explicitly include a reference to the relevant tabular data file, perform these same steps on the next URI template, otherwise use that metadata document.

For example, if the tabular data file is at http://example.org/south-west/devon.csv then processors must attempt to locate a well-known file at http://example.org/.well-known/csvm. If that file contains:

{+url}.json
csvm.json
/csvm?file={url}
        

the processor will first look for http://example.org/south-west/devon.csv.json. If there is no metadata file in that location, it will then look for http://example.org/south-west/csvm.json. Finally, if that also fails, it will look for http://example.org/csvm?file=http://example.org/south-west/devon.csv.json.

If no file were found at http://example.org/.well-known/csvm, the processor will use the default locations and try to retrieve metadata from http://example.org/south-west/devon.csv-metadata.json and, if unsuccessful, http://example.org/south-west/csv-metadata.json.

The use of a well-known location for defining URI patters used to locate metdata files is at risk. The Working Group solicits feedback on whether this mechanism is useful and whether it represents major implementation difficulties.

Embedded Metadata

Most syntaxes for tabular data provide a facility for embedding metadata within the tabular data file itself. The definition of a syntax for tabular data SHOULD include a description of how the syntax maps to an annotated data model, and in particular how any embedded metadata is mapped into the vocabulary defined in [[!tabular-metadata]]. Parsing based on the default dialect for CSV, as described in , will extract column titles from the first row of a CSV file.

        

The results of this can be found in .

For another example, the following tab-delimited file contains embedded metadata where it is assumed that comments may be added using a #, and that the column types may be indicated using a #datatype annotation:

# publisher City of Palo Alto
# updated 12/31/2010
#name GID on_street species trim_cycle  inventory_date
#datatype string  string  string  string  date:M/D/YYYY
  GID On Street Species Trim Cycle  Inventory Date
  1 ADDISON AV  Celtis australis  Large Tree Routine Prune  10/18/2010
  2 EMERSON ST  Liquidambar styraciflua Large Tree Routine Prune  6/2/2010
        

A processor that recognises this format may be able to extract and make sense of this embedded metadata.

Processing Tables

This section describes how particular types of applications should process tabular data and metadata files.

In many cases, an application will start processing from a metadata file. In that case, the initial metadata file is treated as overriding metadata and the application MUST NOT continue to retrieve other available metadata about each of the tabular data files referenced by that initial metadata file other than embedded metadata.

In other cases, applications will start from a tabular data file, such as a CSV file, and locate metadata from that file. This metadata will be used to process the file as if the processor were starting from that metadata file.

For example, if a validator is passed a locally authored metadata file spending.json, which contains:

{
  "tableSchema": "government-spending.csv",
  "tables": [{
    "url": "http://example.org/east-sussex-2015-03.csv",
  }, {
    "url": "http://example.org/east-sussex-2015-02.csv"
  }, ... 
  ]
}
      

the validator would validate all the listed tables, using the locally defined schema at government-spending.csv. It would also use the metadata embedded in the referenced CSV files; for example, when processing http://example.org/east-sussex-2015-03.csv, it would use embedded metadata within that file to verify that the CSV is compatible with the metadata.

If a validator is passed a tabular data file http://example.org/east-sussex-2015-03.csv, the validator would use the metadata located from the CSV file: the first metadata file found through the Link headers found when retrieving that file, or located through a site-wide location configuration.

Starting with a metadata file can remove the need to perform additional requests to locate linked metadata, or metadata retrieved through site-wide location configuration

Creating Annotated Tables

After locating metadata, metadata is normalized and coerced into a single table group description. When starting with a metadata file, this involves normalizing the provided metadata file and verifying that the embedded metadata for each tabular data file referenced from the metadata is compatible with the metadata. When starting with a tabular data file, this involves locating the first metadata file as described in and normalizing into a single descriptor.

If processing starts with a tabular data file, implementations:

  1. Retrieve the tabular data file.
  2. Retrieve the first metadata file (FM) as described in :
    1. metadata supplied by the user (see ).
    2. metadata referenced from a Link Header that may be returned when retrieving the tabular data file (see ).
    3. metadata retrieved through a site-wide location configuration (see ).
    4. embedded metadata as defined in with a single tables entry where the url property is set from that of the tabular data file.
  3. Proceed as if the process starts with FM.

If the process starts with a metadata file:

  1. Retrieve the metadata file yielding the metadata UM (which is treated as overriding metadata, see ).
  2. Normalize UM using the process defined in Normalization in [[!tabular-metadata]], coercing UM into a table group description, if necessary.
  3. For each table (TM) in UM in order, create one or more annotated tables:
    1. Extract the dialect description (DD) from UM for the table associated with the tabular data file. If there is no such dialect description, extract the first available dialect description from a group of tables in which the tabular data file is described. Otherwise use the default dialect description.
    2. If using the default dialect description, override default values in DD based on HTTP headers found when retrieving the tabular data file:
      • If the media type from the Content-Type header is text/tsv, set delimiter to TAB in DD.
      • If the Content-Type header includes the header parameter with a value of absent, set header to false in DD.
      • If the Content-Type header includes the charset parameter, set encoding to this value in DD.
    3. Parse the tabular data file, using DD as a guide, to create a basic tabular data model (T) and extract embedded metadata (EM), for example from the header line.

      This specification provides a non-normative definition for parsing CSV-based files, including the extraction of embedded metadata, in . This specification does not define any syntax for embedded metadata beyond this; whatever syntax is used, it's assumed that metadata can be mapped to the vocabulary defined in [[!tabular-metadata]].

    4. If a Content-Language HTTP header was found when retrieving the tabular data file, and the value provides a single language, set the lang inherited property to this value in TM, unless TM already has a lang inherited property.
    5. Verify that TM is compatible with EM using the procedure defined in Table Description Compatibility in [[!tabular-metadata]]; if TM is not compatible with EM validators MUST raise an error, other processors MUST generate a warning and continue processing.
    6. Use the metadata TM to add annotations to the tabular data model T as described in Section 2 Annotating Tables in [[!tabular-metadata]].

Metadata Compatibility

When processing a tabular data file using metadata as discovered using , processors MUST ensure that the metadata and tabular data file are compatible, this is typically done by extracting embedded metadata from the tabular data file and determining that the provided or discovered metadata is compatible with the embedded metadata using the procedure defined in Table Compatibility in [[!tabular-metadata]].

URL Normalization

Metadata Discovery and Compatibility involve comparing URLs. When comparing URLs, processors MUST use Syntax-Based Normalization as defined in [[!RFC3968]]. Processors MUST perform Scheme-Based Normalization for HTTP (80) and HTTPS (443) and SHOULD perform Scheme-Based Normalization for other well-known schems.

What is generally considered good practice for this in other specifications?

Parsing Cells

Unlike many other data formats, tabular data is designed to be read by humans. For that reason, it's common for data to be represented within tabular data in a human-readable way. The datatype, default, lang, null, required, and separator annotations provide the information needed to parse the string value of a cell into its (semantic) value annotation. This is used:

The process of parsing a cell creates a cell with annotations based on the original string value, parsed value and other column annotations and adds the cell to the list of cells in a row and cells in a column:

After parsing, the cell value can be:

The process of parsing the string value into a single value or a list of values is as follows:

  1. unless the datatype base is string, json, xml, html or anyAtomicType, replace all carriage return (#xD), line feed (#xA), and tab (#x9) characters with space characters.
  2. unless the datatype base is string, json, xml, html, anyAtomicType, or normalizedString, strip leading and trailing whitespace from the string value and replace all instances of two or more whitespace characters with a single space character.
  3. if the normalized string is an empty string, apply the remaining steps to the string given by the column default annotation.
  4. if the column separator annotation is not null and the normalized string is an empty string, the cell value is an empty list. If the column required annotation is true, add an error to the list of errors for the cell.
  5. if the column separator annotation is not null, the cell value is a list of values; set the list annotation on the cell to true, and create the cell value created by:
    1. if the normalized string is the same as any one of the values of the column null annotation, then the resulting value is null.
    2. split the normalized string at the character specified by the column separator annotation.
    3. unless the datatype base is string or anyAtomicType, strip leading and trailing whitespace from these strings.
    4. applying the remaining steps to each of the strings in turn.
  6. if the string is an empty string, apply the remaining steps to the string given by the column default annotation.
  7. if the string is the same as any one of the values of the column null annotation, then the resulting value is null. If the column separator annotation is null and the column required annotation is true, add an error to the list of errors for the cell.
  8. parse the string using the datatype format if one is specified, as described below to give a value with an associated datatype. If the datatype base is string, or there is no datatype, the value has an associated language from the column lang annotation. If there are any errors, add them to the list of errors for the cell; in this case the value has a datatype of string; if the datatype base is string, or there is no datatype, the value has an associated language from the column lang annotation.
  9. validate the value based on the length constraints described in , the value constraints described in and the datatype format annotation if one is specified, as described below. If there are any errors, add them to the list of errors for the cell.

The final value (or values) become the value annotation on the cell.

If there is a about URL annotation on the column, it becomes the about URL annotation on the cell, after being transformed into an absolute URL as described in URI Template Properties of [[!tabular-metadata]].

If there is a property URL annotation on the column, it becomes the property URL annotation on the cell, after being transformed into an absolute URL as described in URI Template Properties of [[!tabular-metadata]].

If there is a value URL annotation on the column, it becomes the value URL annotation on the cell, after being transformed into an absolute URL as described in URI Template Properties of [[!tabular-metadata]]. The value URL annotation is null if the cell value is null and the column virtual annotation is false.

Parsing examples

When datatype annotation is available, the value of a cell is the same as its string value. For example, a cell with a string value of "99" would similarly have the (semantic) value "99".

If a datatype base is provided for the cell, that is used to create a (semantic) value for the cell. For example, if the metadata contains:

  "datatype": "integer"
          

for the cell with the string value "99" then the value of that cell will be the integer 99. A cell whose string value was not a valid integer (such as "one" or "1.0") would be assigned that string value as its (semantic) value annotation, but also have a validation error listed in its errors annotation.

Sometimes data uses special codes to indicate unknown or null values. For example, a particular column might contain a number that is expected to be between 1 and 10, with the string 99 used in the original tabular data file to indicate a null value. The metadata for such a column would include:

  "datatype": {
    "base": "integer",
    "minimum": 1,
    "maximum": 10
  },
  "null": "99"
          

In this case, a cell with a string value of "5" would have the (semantic) value of the integer 5; a cell with a string value of "99" would have the value null.

Similarly, a cell may be assigned a default value if the string value for the cell is empty. A configuration such as:

  "datatype": {
    "base": "integer",
    "minimum": 1,
    "maximum": 10
  },
  "default": "5"
          

In this case, a cell whose string value is "" would be assigned the value of the integer 5. A cell whose string value contains whitespace, such as a single tab character, would also be assigned the value of the integer 5: when the datatype is something other than string or anyAtomicType, leading and trailing whitespace is stripped from string values before the remainder of the processing is carried out.

Cells can contain sequences of values. For example, a cell might have the string value "1 5 7.0". In this case, the separator is a space character. The appropriate configuration would be:

  "datatype": {
    "base": "integer",
    "minimum": 1,
    "maximum": 10
  },
  "default": "5",
  "separator": " "
          

and this would mean that the cell's value would be an array containing two integers and a string: [1, 5, "7.0"]. The final value of the array is a string because it is not a valid integer; the cell's errors annotation will also contain a validation error.

Also, with this configuration, if the string value of the cell were "" (i.e. it was an empty cell) the value of the cell would be an empty list.

A cell value can be inserted into a URL created using a URI template property such as valueUrl. For example, if a cell with the string value "1 5 7.0" were in a column named values, defined with:

  "datatype": "decimal",
  "separator": " ",
  "valueUrl": "{?values}"
          

then after expansion of the URI template, the resulting valueUrl would be ?values=1.0,5.0,7.0. The canonical representations of the decimal values are used within the URL.

Formats for numeric types

It is not uncommon for numbers within tabular data to be formatted for human consumption, which may involve using commas for decimal points, grouping digits in the number using commas, or adding currency symbols or percent signs to the number.

If the datatype base is a numeric type, the datatype format annotation indicates the expected format for that number. Its value MUST be either a single string or an object with one or more of the properties:

decimalChar
A string whose value is used to represent a decimal point within the number. The default value is ".".
groupChar
A string whose value is used to group digits within the number. The default value is ",".
pattern
A regular expression string, with syntax and processing defined by [[!ECMASCRIPT]].

Authors are encouraged to be conservative in the regular expressions that they use, sticking to the basic features of regular expressions that are likely to be supported across implementations.

If the datatype format annotation is a single string, this is interpreted in the same way as if it were an object with a pattern property whose value is that string.

When parsing the string value of a cell against this format specification, implementations MUST recognise and parse numbers that consist of:

  1. an optional + or - sign,
  2. followed by a decimal digit (0-9),
  3. followed by any number of decimal digits (0-9) and the string specified as the groupChar,
  4. followed by an optional decimalChar followed by one or more decimal digits (0-9),
  5. followed by an optional exponent, consisting of an E followed by an optional + or - sign followed by one or more decimal digits (0-9), or
  6. followed by an optional percent (%) or per-mille () sign.

or that are one of the special values:

  1. NaN,
  2. INF, or
  3. -INF.

Implementations MUST add a validation error to the errors annotation for the cell if the string being parsed:

  • does not meet the numeric format defined above,
  • contains two consecutive groupChar strings,
  • does not match the regular expression defined in the pattern property, if there is one,
  • contains the decimalChar, if the datatype base is integer or one of its sub-values,
  • contains an exponent, if the datatype base is decimal or one of its sub-values, or
  • is one of the special values NaN, INF, or -INF, if the datatype base is decimal or one of its sub-values.

Implementations MUST use the sign, exponent, percent, and per-mille signs when parsing the string value of a cell to provide the value of the cell. For example, the string value "-25%" must be interpreted as -0.25 and the string value "1E6" as 1000000.

Formats for booleans

Boolean values may be represented in many ways aside from the standard 1 and 0 or true and false.

If the datatype base for a cell is boolean, the datatype format annotation provides the true value followed by the false value, separated by |. For example if format is Y|N then cells must hold either Y or N with Y meaning true and N meaning false.

The resulting cell value will be one or more boolean true or false values.

Formats for dates and times

Dates and times are commonly represented in tabular data in formats other than those defined in [[!xmlschema11-2]].

If the datatype base is a date or time type, the datatype format annotation indicates the expected format for that date or time.

The supported date and time formats listed here are expressed in terms of the date field symbols defined in [[!UAX35]] and MUST be interpreted by implementations as defined in that specification.

The following date formats MUST be recognized by implementations:

  • yyyy-MM-dd e.g., 2015-03-22
  • yyyyMMdd e.g., 20150322
  • dd-MM-yyyy e.g., 22-03-2015
  • d-M-yyyy e.g., 22-3-2015
  • MM-dd-yyyy e.g., 03-22-2015
  • M-d-yyyy e.g., 3-22-2015
  • dd/MM/yyyy e.g., 22/03/2015
  • d/M/yyyy e.g., 22/3/2015
  • MM/dd/yyyy e.g., 03/22/2015
  • M/d/yyyy e.g., 3/22/2015
  • dd.MM.yyyy e.g., 22.03.2015
  • d.M.yyyy e.g., 22.3.2015
  • MM.dd.yyyy e.g., 03.22.2015
  • M.d.yyyy e.g., 3.22.2015

The following time formats MUST be recognized by implementations:

  • HH:mm:ss.S with one or more trailing S characters indicating the maximum number of fractional seconds e.g., HH:mm:ss.SSS for 15:02:37.143
  • HH:mm:ss e.g., 15:02:37
  • HHmmss e.g., 150237
  • HH:mm e.g., 15:02
  • HHmm e.g., 1502

The following date/time formats MUST be recognized by implementations:

  • yyyy-MM-ddTHH:mm:ss.S with one or more trailing S characters indicating the maximum number of fractional seconds e.g., yyyy-MM-ddTHH:mm:ss.SSS for 2015-03-15T15:02:37.143
  • yyyy-MM-ddTHH:mm:ss e.g., 2015-03-15T15:02:37
  • yyyy-MM-ddTHH:mm e.g., 2015-03-15T15:02
  • any of the date formats above, followed by a single space, followed by any of the time formats above, e.g., M/d/yyyy HH:mm for 3/22/2015 15:02 or dd.MM.yyyy HH:mm:ss for 22.03.2015 15:02:37

Implementations MUST also recognise date, time, and date/time formats that end with timezone markers consisting of between one and three x or X characters, possibly after a single space. These MUST be interpreted as follows:

  • X e.g., -08, +0530, or Z (minutes are optional)
  • XX e.g., -0800, +0530, or Z
  • XXX e.g., -08:00, +05:30, or Z
  • x e.g., -08 or +0530 (Z is not permitted)
  • xx e.g., -0800 or +0530 (Z is not permitted)
  • xxx e.g., -08:00 or +05:30 (Z is not permitted)

For example, formats could include yyyy-MM-ddTHH:mm:ssXXX for 2015-03-15T15:02:37Z or 2015-03-15T15:02:37-05:00, or HH:mm x for 15:02 -05.

The cell value will one or more dates/time values extracted using the format.

For simplicity, this version of this standard does not support abbreviated or full month or day names, or double digit years. Future versions of this standard may support other date and time formats, or general purpose date/time pattern strings. Authors of schemas SHOULD use appropriate regular expressions, along with the string datatype, for dates and times that use a format other than that specified here.

Formats for durations

Durations MUST be formatted and interpreted as defined in [[!xmlschema11-2]], using the [[!ISO8601]] format -?PnYnMnDTnHnMnS. For example, the duration P1Y1D is used for a year and a day; the duration PT2H30M for 2 hours and 30 minutes.

If the datatype base is a duration type, the datatype format annotation provides a regular expression for the string values, with syntax and processing defined by [[!ECMASCRIPT]].

Authors are encouraged to be conservative in the regular expressions that they use, sticking to the basic features of regular expressions that are likely to be supported across implementations.

The cell value will be one or more durations extracted using the format.

Formats for other types

If the datatype base is not numeric, boolean, a date/time type, or a duration type, the datatype format annotation provides a regular expression for the string values, with syntax and processing defined by [[!ECMASCRIPT]].

Authors are encouraged to be conservative in the regular expressions that they use, sticking to the basic features of regular expressions that are likely to be supported across implementations.

Values that are labelled as html, xml, or json SHOULD NOT be validated against those formats.

Metadata creators who wish to check the syntax of HTML, XML, or JSON within tabular data should use the datatype format annotation to specify a regular expression against which such values will be tested.

Presenting Tables

When presenting tables, implementations should:

Bidirectional Tables

There are two levels of bidirectionality to consider when displaying tables: the directionality of the table (i.e., whether the columns should be arranged left-to-right or right-to-left) and the directionality of the content of individual cells.

The direction annotation on the table provides information about the desired display of the table. If direction is ltr then the first column should be displayed on the left and the last column on the right. If direction is rtl then the first column should be displayed on the right and the last column on the left.

If direction is auto then tables should be displayed with attention to the bidirectionality of the content of the file. Specifically, the values of the cells in the table should be scanned breadth first: from the first cell in the first column through to the last cell in the first row, down to the last cell in the last column. If the first character in the table with a strong type as defined in [[!BIDI]] indicates a RTL directionality, the table should be displayed with the first column on the right and the last column on the left. Otherwise, the table should be displayed with the first column on the left and the last column on the right. Characters such as whitespace, quotes, commas, and numbers do not have a strong type, and therefore are skipped when identifying the character that determines the directionality of the table.

Implementations should enable user preferences to override the indicated metadata about the directionality of the table.

Once the directionality of the table has been determined, each cell within the table should be considered as a separate paragraph, as defined by the Unicode Bidirectional Algorithm (UBA) in [[!BIDI]]. The default directionality for the cell is determined by looking at the text direction annotation for the cell.

Thus, as defined by the UBA, if a cell contains no characters with a strong type (if it's a number or date for example) then the way the cell is displayed should be determined by the text direction annotation of the cell. However, when the cell contains characters with a strong type (such as letters) then they should be displayed according to the Unicode Bidirectional Algorithm as described in [[!BIDI]].

Column and row labelling

The labelling of columns and rows helps those who are attempting to understand the content of a table to grasp what a particular cell means. Implementations should present appropriate titles for columns, and ensure that the most important information in a row is kept apparent to the user, to aid their understanding. For example:

  • a table presented on the screen might retain certain columns in view so that readers can easily glance at the identifying information in each row
  • as the user moves focus into a cell, screen readers announce a label for the new column if the user has changed column, or for the new row if the user has changed row

When labelling a column, either on the screen or aurally, implementations should use the first available of:

  1. the column's titles in the preferred language of the user, or with an undefined language if there is no title available in a preferred language; there may be multiple such titles in which case all should be announced
  2. the column's name
  3. the column's number

When labelling a row, either on the screen or aurally, implementations should use the first available of:

  1. the row's titles in the preferred language of the user, or with an undefined language if there is no title available in a preferred language; there may be multiple such titles in which case all should be announced
  2. the values of the cells in the row's primary key
  3. the row's number

Validating Tables

Validators test whether given tabular data files adhere to the structure defined within a schema. Validators MUST raise errors as defined in [[!tabular-metadata]] and in the following situations:

Processors MAY continue validating after encountering the first error.

Converting Tables

Conversions of tabular data to other formats operate over a annotated table constructed as defined in Annotating Tables in [[!tabular-metadata]]. The mechanics of these conversions to other formats are defined in other specifications such as [[csv2json]] and [[csv2rdf]].

Conversion specifications MUST define a default mapping from an annotated table that lacks any annotations (i.e., that is equivalent to an un-annotated table).

Conversion specifications MUST use the property value of the propertyUrl of a column as the basis for naming machine-readable fields in the target format, such as the name of the equivalent element or attribute in XML, property in JSON or property URI in RDF.

Conversion specifications MAY use any of the annotations found on an annotated table group, table, column, row or cell, including non-core annotations, to adjust the mapping into another format.

Conversion specifications MAY define additional annotations, not defined in this specification, which are specifically used when converting to the target format of the conversion. For example, a conversion to XML might specify a http://example.org/conversion/xml/element-or-attribute property on columns that determines whether a particular column is represented through an element or an attribute in the data.

Best Practice CSV

There is no standard for CSV, and there are many variants of CSV used on the web today. This section defines a method for expressing tabular data adhering to the annotated tabular data model in CSV. Authors are encouraged to adhere to the constraints described in this section as implementations should process such CSV files consistently.

This syntax is not compliant with text/csv as defined in [[!RFC4180]] in that it permits line endings other than CRLF. Supporting LF line endings is important for data formats that are used on non-Windows platforms. However, all files that adhere to [[!RFC4180]]'s definition of CSV meet the constraints described in this section.

Developing a standard for CSV is outside the scope of the Working Group. The details here aim to help shape any future standard.

Content Type

The appropriate content type for a CSV file is text/csv. For example, when a CSV file is transmitted via HTTP, the HTTP response should include a Content-Type header with the value text/csv:

Content-Type: text/csv
        

Encoding

CSV files should be encoded using UTF-8. If a CSV file is not encoded using UTF-8, the encoding should be specified through the charset parameter in the Content-Type header:

Content-Type: text/csv;charset=ISO-8859-1
        

Line Endings

The ends of rows in a CSV file should be CRLF (U+000D U+000A) but may be LF (U+000A). Line endings within escaped cells are not normalised.

Lines

Each line of a CSV file should contain the same number of comma-separated values.

Values that contain commas, line endings, or double quotes should be escaped by having the entire value wrapped in double quotes. There should not be whitespace before or after the double quotes. Within these escaped cells, any double quotes should be escaped with two double quotes ("").

Headers

The first line of a CSV file should contain a comma-separated list of names of columns. This is known as the header line and provides titles for the columns. There are no constraints on these titles.

If a CSV file does not include a header line, this should be specified using the header parameter of the media type:

Content-Type: text/csv;header=absent
          

Grammar

This grammar is a generalization of that defined in [[!RFC4180]] and is included for reference only.

The EBNF used here is defined in XML 1.0 [[EBNF-NOTATION]].

Parsing Tabular Data

As described in , there may be many formats which an application might interpret into the tabular data model described in , including using different separators or fixed format tables, multiple tables within a single file, or ones that have metadata lines before a table header.

Standardizing the parsing of CSV is outside the chartered scope of the Working Group. This non-normative section is intended to help the creators of parsers handle the wide variety of CSV-based formats that they may encounter due to the current lack of standardization of the format.

This section describes an algorithm for parsing formats that do not adhere to the constraints described in , as well as those that do, and extracting embedded metadata. The parsing algorithm uses the following flags. These may be set by metadata properties found while Locating Metadata, including through user input (see Overriding Metadata), or through the inclusion of a dialect description within a metadata file:

comment prefix
A string that, when it appears at the beginning of a row, indicates that the row is a comment that should be associated as a rdfs:comment annotation to the table. This is set by the commentPrefix property of a dialect description. The default is null, which means no rows are treated as comments. A value other than null may mean that the source numbers of rows are different from their numbers.
delimiter
The separator between cells, set by the delimiter property of a dialect description. The default is ,.
encoding
The character encoding for the file, one of the encodings listed in [[encoding]], set by the encoding property of a dialect description. The default is utf-8.
escape character
The string that is used to escape the quote character within escaped cells, or null, set by the doubleQuote property of a dialect description. The default is " (such that "" is used to escape " within an escaped cell).
header row count
The number of header rows (following the skipped rows) in the file, set by the header or headerRowCount property of a dialect description. The default is 1. A value other than 0 will mean that the source numbers of rows will be different from their numbers.
line terminators
The strings that can be used at the end of a row, set by the lineTerminators property of a dialect description. The default is [CRLF, LF].
quote character
The string that is used around escaped cells, or null, set by the quoteChar property of a dialect description. The default is ".
skip blank rows
Indicates whether to ignore wholly empty rows (i.e. rows in which all the cells are empty), set by the skipBlankRows property of a dialect description. The default is false. A value other than false may mean that the source numbers of rows are different from their numbers.
skip columns
The number of columns to skip at the beginning of each row, set by the skipColumns property of a dialect description. The default is 0. A value other than 0 will mean that the source numbers of columns will be different from their numbers.
skip rows
The number of rows to skip at the beginning of the file, before a header row or tabular data, set by the skipRows property of a dialect description. The default is 0. A value greater than 0 will mean that the source numbers of rows will be different from their numbers.
trim
Indicates whether to trim whitespace around cells; may be true, false, start, or end, set by the skipInitialSpace or trim property of a dialect description. The default is true.

The algorithm for using these flags to parse a document containing tabular data to create a basic annotated tabular data model and to extract embedded metadata is as follows:

  1. Create a new table T with the annotations:
  2. Create a metadata document structure M that looks like:
    {
      "@context": "http://www.w3.org/ns/csvw",
      "rdfs:comment": []
      "tableSchema": {
        "columns": []
      }
    }
              
  3. If the URL of the tabular data file being parsed is known, set the url property on M to that URL.
  4. Set source row number to 1.
  5. Read the file using the encoding, as specified in [[encoding]], using the replacement error mode. If the encoding is not a Unicode encoding, use a normalizing transcoder to normalize into Unicode Normal Form C as defined in [[UAX15]].

    The replacement error mode ensures that any non-Unicode characters within the CSV file are replaced by U+FFFD, ensuring that strings within the tabular data model such as column titles and cell string values only contain valid Unicode characters.

  6. Repeat the following the number of times indicated by skip rows:
    1. Read a row to provide the row content.
    2. If the comment prefix is not null and the row content begins with the comment prefix, strip that prefix from the row content, and add the resulting string to the M.rdfs:comment array.
    3. Otherwise, if the row content is not an empty string, add the row content to the M.rdfs:comment array.
    4. Add 1 to the source row number.
  7. Repeat the following the number of times indicated by header row count:
    1. Read a row to provide the row content.
    2. If the comment prefix is not null and the row content begins with the comment prefix, strip that prefix from the row content, and add the resulting string to the M.rdfs:comment array.
    3. Otherwise, parse the row to provide a list of cell values, and:
      1. Remove the first skip columns number of values from the list of cell values.
      2. For each of the remaining values at index i in the list of cell values:
        1. If the value at index i in the list of cell values is an empty string or consists only of whitespace, do nothing.
        2. Otherwise, if there is no column description object at index i in M.tableSchema.columns, create a new one with a title property whose value is an array containing a single value that is the value at index i in the list of cell values.
        3. Otherwise, add the value at index i in the list of cell values to the array at M.tableSchema.columns[i].title.
    4. Add 1 to the source row number.
  8. If header row count is zero, create an empty column description object in M.tableSchema.columns for each column in the current row after skip columns.
  9. Set row number to 1.
  10. While it is possible to read another row, do the following:
    1. Set the source column number to 1.
    2. Read a row to provide the row content.
    3. If the comment prefix is not null and the row content begins with the comment prefix, strip that prefix from the row content, and add the resulting string to the M.rdfs:comment array.
    4. Otherwise, parse the row to provide a list of cell values, and:
      1. If all of the values in the list of cell values are empty strings, and skip blank rows is true, add 1 to the source row number and move on to process the next row.
      2. Otherwise, create a new row R, with:
      3. Append R to the rows of table T.
      4. Remove the first skip columns number of values from the list of cell values and add that number to the source column number.
      5. For each of the remaining values at index i in the list of cell values (where i starts at 1):
        1. Identify the column C at index i within the columns of table T. If there is no such column:
          1. Create a new column C with:
          2. Append C to the columns of table T (at index i).
        2. Create a new cell D, with:
        3. Append cell D to the cells of column C.
        4. Append cell D to the cells of row R (at index i).
        5. Add 1 to the source column number.
    5. Add 1 to the source row number.
  11. If M.rdfs:comment is an empty array, remove the rdfs:comment property from M.
  12. Return the table T and the embedded metadata M.

To read a row to provide row content, perform the following steps:

  1. Set the row content to an empty string.
  2. Read initial characters and process as follows:
    1. If the string starts with the escape character followed by the quote character, append both strings to the row content, and move on to process the string following the quote character.
    2. Otherwise, if the string starts with the escape character and the escape character is not the same as the quote character, append the escape character and the single character following it to the row content and move on to process the string following that character.
    3. Otherwise, if the string starts with the quote character, append the quoted value obtained by reading a quoted value to the row content and move on to process the string following the quoted value.
    4. Otherwise, if the string starts with one of the line terminators, return the row content.
    5. Otherwise, append the first character to the row content and move on to process the string following that character.
  3. If there are no more characters to read, return the row content.

To read a quoted value to provide a quoted value, perform the following steps:

  1. Set the quoted value to an empty string.
  2. Read the initial quote character and add a quote character to the quoted value.
  3. Read initial characters and process as follows:
    1. If the string starts with the escape character followed by the quote character, append both strings to the quoted value, and move on to process the string following the quote character.
    2. Otherwise, if string starts with the escape character and the escape character is not the same as the quote character, append the escape character and the character following it to the quoted value and move on to process the string following that character.
    3. Otherwise, if the string starts with the quote character, return the quoted value.
    4. Otherwise, append the first character to the quoted value and move on to process the string following that character.

To parse a row to provide a list of cell values, perform the following steps:

  1. Set the list of cell values to an empty list and the current cell value to an empty string.
  2. Set the quoted flag to false.
  3. Read initial characters and process as follows:
    1. If the string starts with the escape character followed by the quote character, append the quote character to the current cell value, and move on to process the string following the quote character.
    2. Otherwise, if the string starts with the escape character and the escape character is not the same as the quote character, append the character following the escape character to the current cell value and move on to process the string following that character.
    3. Otherwise, if the string starts with the quote character then:
      1. If quoted is false, set the quoted flag to true, and move on to process the remaining string. If the current cell value is not an empty string, raise an error.
      2. Otherwise, set quoted to false, and move on to process the remaining string. If the remaining string does not start with the delimiter, raise an error.
    4. Otherwise, if the string starts with the delimiter, then:
      1. If quoted is true, append the delimiter string to the current cell value and move on to process the remaining string.
      2. Otherwise, conditionally trim the current cell value, add the resulting trimmed cell value to the list of cell values and move on to process the following string.
    5. Otherwise, append the first character to the current cell value and move on to process the remaining string.
  4. If there are no more characters to read, conditionally trim the current cell value, add the resulting trimmed cell value to the list of cell values and return the list of cell values.

To conditionally trim a cell value to provide a trimmed cell value, perform the following steps:

  1. Set the trimmed cell value to the provided cell value.
  2. If trim is true or start then remove any leading whitespace from the start of the trimmed cell value and move on to the next step.
  3. If trim is true or end then remove any trailing whitespace from the end of the trimmed cell value and move on to the next step.
  4. Return the trimmed cell value.

This parsing algorithm does not account for the possibility of there being more than one area of tabular data within a single CSV file.

Bidirectionality in CSV Files

Bidirectional content does not alter the definition of rows or the assignment of cells to columns. Whether or not a CSV file contains right-to-left characters, the first column's content is the first cell of each row, which is the text prior to the first occurrence of a comma within that row.

For example, Egyptian Referendum results are available as a CSV file at https://egelections-2011.appspot.com/Referendum2012/results/csv/EG.csv. Over the wire and in non-Unicode-aware text editors, the CSV looks like:

            
‌ا‌ل‌م‌ح‌ا‌ف‌ظ‌ة‌,‌ن‌س‌ب‌ة‌ ‌م‌و‌ا‌ف‌ق‌,‌ن‌س‌ب‌ة‌ ‌غ‌ي‌ر‌ ‌م‌و‌ا‌ف‌ق‌,‌ع‌د‌د‌ ‌ا‌ل‌ن‌ا‌خ‌ب‌ي‌ن‌,‌ا‌ل‌أ‌ص‌و‌ا‌ت‌ ‌ا‌ل‌ص‌ح‌ي‌ح‌ة‌,‌ا‌ل‌أ‌ص‌و‌ا‌ت‌ ‌ا‌ل‌ب‌ا‌ط‌ل‌ة‌,‌ن‌س‌ب‌ة‌ ‌ا‌ل‌م‌ش‌ا‌ر‌ك‌ة‌,‌م‌و‌ا‌ف‌ق‌,‌غ‌ي‌ر‌ ‌م‌و‌ا‌ف‌ق‌
‌ا‌ل‌ق‌ل‌ي‌و‌ب‌ي‌ة‌,60.0,40.0,"2,639,808","853,125","15,224",32.9,"512,055","341,070"
‌ا‌ل‌ج‌ي‌ز‌ة‌,66.7,33.3,"4,383,701","1,493,092","24,105",34.6,"995,417","497,675"
‌ا‌ل‌ق‌ا‌ه‌ر‌ة‌,43.2,56.8,"6,580,478","2,254,698","36,342",34.8,"974,371","1,280,327"
‌ق‌ن‌ا‌,84.5,15.5,"1,629,713","364,509","6,743",22.8,"307,839","56,670"
...
            
          

Within this CSV file, the first column appears as the content of each line before the first comma and is named المحافظة (appearing at the start of each row as ‌ا‌ل‌م‌ح‌ا‌ف‌ظ‌ة‌ in the example, which is displaying the relevant characters from left to right in the order they appear "on the wire").

The CSV translates to a table model that looks like:

Column / Row column 1 column 2 column 3 column 4 column 5 column 6 column 7 column 8 column 9
column namesالمحافظةنسبة موافقنسبة غير موافقعدد الناخبينالأصوات الصحيحةالأصوات الباطلةنسبة المشاركةموافقغير موافق
row 1القليوبية60.040.02,639,808853,12515,22432.9512,055341,070
row 2الجيزة66.733.34,383,7011,493,09224,10534.6995,417497,675
row 3القاهرة43.256.86,580,4782,254,69836,34234.8974,3711,280,327
row 4قنا84.515.51,629,713364,5096,74322.8307,83956,670

The fragment identifier #col=3 identifies the third of the columns, named نسبة غير موافق (appearing as ‌ن‌س‌ب‌ة‌ ‌غ‌ي‌ر‌ ‌م‌و‌ا‌ف‌ق‌ in the example).

defines how this table model should be displayed by compliant applications, and how metadata can affect the display. The default is for the display to be determined by the content of the table. For example, if this CSV were turned into an HTML table for display into a web page, it should be displayed with the first column on the right and the last on the left, as follows:

غير موافق موافق نسبة المشاركة الأصوات الباطلة الأصوات الصحيحة عدد الناخبين نسبة غير موافق نسبة موافق المحافظة
341,070 512,055 32.9 15,224 853,125 2,639,808 40.0 60.0 القليوبية
497,675 995,417 34.6 24,105 1,493,092 4,383,701 33.3 66.7 الجيزة
1,280,327 974,371 34.8 36,342 2,254,698 6,580,478 56.8 43.2 القاهرة
56,670 307,839 22.8 6,743 364,509 1,629,713 15.5 84.5 قنا

The fragment identifier #col=3 still identifies the third of the columns, named نسبة غير موافق, which appears in the HTML display as the third column from the right and is what those who read right-to-left would think of as the third column.

Note that this display matches that shown on the original website.

An alternative approach is for the CSV to be parsed into a table model in which the columns are numbered in the reverse, for tables which are either marked as or detected to be right-to-left tables. For example, we could introduce a bidi=rtl or similar media type parameter, and use this to determine whether the first column in table generated from the CSV is the text before the first comma in each line or the text after the last comma in the line.

In the example above, if the CSV were served with bidi=rtl, or the table was detected as being a right-to-left table, then the column numbering in the model would be reversed:

Column / Row column 9 column 8 column 7 column 6 column 5 column 4 column 3 column 2 column 1
row 1 (header)المحافظةنسبة موافقنسبة غير موافقعدد الناخبينالأصوات الصحيحةالأصوات الباطلةنسبة المشاركةموافقغير موافق
row 2القليوبية60.040.02,639,808853,12515,22432.9512,055341,070
row 3الجيزة66.733.34,383,7011,493,09224,10534.6995,417497,675
row 4القاهرة43.256.86,580,4782,254,69836,34234.8974,3711,280,327
row 5قنا84.515.51,629,713364,5096,74322.8307,83956,670

This would require a change to [[RFC7111]] but that might be required by updates to the definition of text/csv in any case. With the change, the fragment identifier #col=3 would then refer to the third column from the right, named نسبة المشاركة.

If the model were defined in this way, there would be no need to determine the order of the columns when displayed using a metadata property. Columns would always be displayed with the first column (numbered 1 in the model) on the left. The final display in HTML, for example, would be exactly as above. The only difference would be that #col=3 would refer to the third column from the left.

We note that using media type parameters is problematic because publishers might not have the ability to set them on their servers, and because they can easily get lost as a file is republished or emailed between people.

We invite comment on the best way to approach bidirectionality in CSV files.

Examples

Simple Example

A simple CSV file that complies to the constraints described in , at http://example.org/tree-ops.csv, might look like:

          

Parsing this file results in an annotated tabular data model of a single table T with five columns and two rows. The columns have the annotations shown in the following table:

idcore annotations
tablenumbersource numbercellstitles
C1T11C1.1, C2.1GID
C2T22C1.2, C2.2On Street
C3T33C1.3, C2.3Species
C4T44C1.4, C2.4Trim Cycle
C5T55C1.5, C2.5Inventory Date

The extracted embedded metadata, as defined in [[!tabular-metadata]], would look like:

{
  "@type": "Table",
  "url": "http://example.org/tree-ops.csv",
  "tableSchema": {
    "columns": [
      {"titles": [ "GID" ]},
      {"titles": [ "On Street" ]},
      {"titles": [ "Species" ]},
      {"titles": [ "Trim Cycle" ]},
      {"titles": [ "Inventory Date" ]}
    ]
  }
}
          

The rows have the annotations shown in the following table:

idcore annotations
tablenumbersource numbercells
R1T12C1.1, C1.2, C1.3, C1.4, C1.5
R2T23C2.1, C2.2, C2.3, C2.4, C2.5

The source number of each row is offset by one from the number of each row because in the source CSV file, the header line is the first line. It is possible to reconstruct a [[RFC7111]] compliant reference to the first record in the original CSV file (http://example.org/tree-ops.csv#row=2) using the value of the row's source number. This enables implementations to retain provenance between the table model and the original file.

The cells have the annotations shown in the following table (note that the values of all the cells in the table are strings, denoted by the double quotes in the table below):

idcore annotations
tablecolumnrowstring valuevalue
C1.1TC1R1"1""1"
C1.2TC2R1"ADDISON AV""ADDISON AV"
C1.3TC3R1"Celtis australis""Celtis australis"
C1.4TC4R1"Large Tree Routine Prune""Large Tree Routine Prune"
C1.5TC5R1"10/18/2010""10/18/2010"
C2.1TC1R2"2""2"
C2.2TC2R2"EMERSON ST""EMERSON ST"
C2.3TC3R2"Liquidambar styraciflua""Liquidambar styraciflua"
C2.4TC4R2"Large Tree Routine Prune""Large Tree Routine Prune"
C2.5TC5R2"6/2/2010""6/2/2010"
Using Overriding Metadata

The tools that the consumer of this data uses may provide a mechanism for overriding the metadata that has been provided within the file itself. For example, they might enable the consumer to add machine-readable names to the columns, or to mark the fifth column as holding a date in the format M/D/YYYY. These facilities are implementation defined; the code for invoking a Javascript-based parser might look like:

data.parse({
  "column-names": ["GID", "on_street", "species", "trim_cycle", "inventory_date"],
  "datatypes": ["string", "string", "string", "string", "date"],
  "formats": [null,null,null,null,"M/D/YYYY"]
});
          	

This is equivalent to a metadata file expressed in the syntax defined in [[!tabular-metadata]], looking like:

{
  "@type": "Table",
  "url": "http://example.org/tree-ops.csv",
  "tableSchema": {
    "columns": [{
      "name": "GID",
      "datatype": "string"
    }, {
      "name": "on_street",
      "datatype": "string"
    }, {
      "name": "species",
      "datatype": "string"
    }, {
      "name": "trim_cycle",
      "datatype": "string"
    }, {
      "name": "inventory_date",
      "datatype": {
        "base": "date",
        "format": "M/d/yyyy"
      }
    }]
  }
}
            

This would be merged with the embedded metadata found in the CSV file, providing the titles for the columns to create:

{
  "@type": "Table",
  "url": "http://example.org/tree-ops.csv",
  "tableSchema": {
    "columns": [{
      "name": "GID", 
      "titles": "GID", 
      "datatype": "string"
    }, {
      "name": "on_street", 
      "titles": "On Street", 
      "datatype": "string"
    }, {
      "name": "species", 
      "titles": "Species", 
      "datatype": "string"
    }, {
      "name": "trim_cycle", 
      "titles": "Trim Cycle", 
      "datatype": "string"
    }, {
      "name": "inventory_date",
      "titles": "Inventory Date",
      "datatype": {
        "base": "date", 
        "format": "M/d/yyyy"
      }
    }]
  }
}
            

The processor can then create an annotated tabular data model that included name annotations on the columns, and datatype annotations on the cells, and created cells whose values were of appropriate types (in the case of this Javascript implementation, the cells in the last column would be Date objects, for example).

Assuming this kind of implementation-defined parsing, the columns would then have the annotations shown in the following table:

idcore annotations
tablenumbersource numbercellsnametitlesdatatype
C1T11C1.1, C2.1GIDGIDstring
C2T22C1.2, C2.2on_streetOn Streetstring
C3T33C1.3, C2.3speciesSpeciesstring
C4T44C1.4, C2.4trim_cycleTrim Cyclestring
C5T55C1.5, C2.5inventory_dateInventory Date{ "base": "date", "format": "M/d/yyyy" }

The cells have the annotations shown in the following table. Because of the overrides provided by the consumer to guide the parsing, and the way the parser works, the cells in the Inventory Date column (cells C1.5 and C2.5) have values that are parsed dates rather than unparsed strings.

idcore annotations
tablecolumnrowstring valuevalue
C1.1TC1R1"1""1"
C1.2TC2R1"ADDISON AV""ADDISON AV"
C1.3TC3R1"Celtis australis""Celtis australis"
C1.4TC4R1"Large Tree Routine Prune""Large Tree Routine Prune"
C1.5TC5R1"10/18/2010"2010-10-18
C2.1TC1R2"2""2"
C2.2TC2R2"EMERSON ST""EMERSON ST"
C2.3TC3R2"Liquidambar styraciflua""Liquidambar styraciflua"
C2.4TC4R2"Large Tree Routine Prune""Large Tree Routine Prune"
C2.5TC5R2"6/2/2010"2010-06-02
Using a Metadata File

A similar set of annotations could be provided through a metadata file, located as discussed in and defined in [[!tabular-metadata]]. For example, this might look like:

            

The annotated tabular data model generated from this would be more sophisticated again. The table itself would have the following annotations:

dc:title
{"@value": "Tree Operations", "@language": "en"}
dcat:keyword
[{"@value": "tree", "@language", "en"}, {"@value": "street", "@language": "en"}, {"@value": "maintenance", "@language": "en"}]
dc:publisher
[{ "schema:name": "Example Municipality", "schema:url": {"@id": "http://example.org"} }]
dc:license
{"@id": "http://opendefinition.org/licenses/cc-by/"}
dc:modified
{"@value": "2010-12-31", "@type": "date"}

The columns would have the annotations shown in the following table:

idcore annotationsother annotations
tablenumbersource numbercellsnametitlesdatatypedc:description
C1T11C1.1, C2.1GIDGID, Generic IdentifierstringAn identifier for the operation on a tree.
C2T22C1.2, C2.2on_streetOn StreetstringThe street that the tree is on.
C3T33C1.3, C2.3speciesSpeciesstringThe species of the tree.
C4T44C1.4, C2.4trim_cycleTrim CyclestringThe operation performed on the tree.
C5T55C1.5, C2.5inventory_dateInventory Date{ "base": "date", "format": "M/d/yyyy" }The date of the operation that was performed.

The rows have an additional primary key annotation, as shown in the following table:

idcore annotations
tablenumbersource numbercellsprimary key
R1T12C1.1, C1.2, C1.3, C1.4, C1.5C1.1
R2T23C2.1, C2.2, C2.3, C2.4, C2.5C2.1

Thanks to the provided metadata, the cells again have the annotations shown in the following table. The metadata file has provided the information to supplement the model with additional annotations but also, for the Inventory Date column (cells C1.5 and C2.5), have a value that is a parsed date rather than an unparsed string.

idcore annotations
tablecolumnrowstring valuevalueabout URL
C1.1TC1R1"1""1"http://example.org/tree-ops.csv#gid-1
C1.2TC2R1"ADDISON AV""ADDISON AV"http://example.org/tree-ops.csv#gid-1
C1.3TC3R1"Celtis australis""Celtis australis"http://example.org/tree-ops.csv#gid-1
C1.4TC4R1"Large Tree Routine Prune""Large Tree Routine Prune"http://example.org/tree-ops.csv#gid-1
C1.5TC5R1"10/18/2010"2010-10-18http://example.org/tree-ops.csv#gid-1
C2.1TC1R2"2""2"http://example.org/tree-ops.csv#gid-2
C2.2TC2R2"EMERSON ST""EMERSON ST"http://example.org/tree-ops.csv#gid-2
C2.3TC3R2"Liquidambar styraciflua""Liquidambar styraciflua"http://example.org/tree-ops.csv#gid-2
C2.4TC4R2"Large Tree Routine Prune""Large Tree Routine Prune"http://example.org/tree-ops.csv#gid-2
C2.5TC5R2"6/2/2010"2010-06-02http://example.org/tree-ops.csv#gid-2

Empty and Quoted Cells

The following slightly amended CSV file contains quoted and missing cell values:

GID,On Street,Species,Trim Cycle,Inventory Date
1,ADDISON AV,"Celtis australis","Large Tree Routine Prune",10/18/2010
2,,"Liquidambar styraciflua","Large Tree Routine Prune",
          

Parsing this file similarly results in an annotated tabular data model of a single table T with five columns and two rows. The columns and rows have exactly the same annotations as previously, but there are two null cell values for C2.2 and C2.5. Note that the quoting of values within the CSV makes no difference to either the string value or value of the cell.

idcore annotations
tablecolumnrowstring valuevalue
C1.1TC1R1"1""1"
C1.2TC2R1"ADDISON AV""ADDISON AV"
C1.3TC3R1"Celtis australis""Celtis australis"
C1.4TC4R1"Large Tree Routine Prune""Large Tree Routine Prune"
C1.5TC5R1"10/18/2010""10/18/2010"
C2.1TC1R2"2""2"
C2.2TC2R2""null
C2.3TC3R2"Liquidambar styraciflua""Liquidambar styraciflua"
C2.4TC4R2"Large Tree Routine Prune""Large Tree Routine Prune"
C2.5TC5R2""null

Tabular Data Embedding Annotations

The following example illustrates some of the complexities that can be involved in parsing tabular data, how the flags described above can be used, and how new tabular data formats could be defined that embed additional annotations into the tabular data model.

In this example, the publishers of the data are using an internal convention to supply additional metadata about the tabular data embedded within the file itself. They are also using a tab as a separator rather than a comma.

#	publisher	City of Palo Alto
#	updated	12/31/2010
#name	GID	on_street	species	trim_cycle	inventory_date
#datatype	string	string	string	string	date:M/D/YYYY
	GID	On Street	Species	Trim Cycle	Inventory Date
	1	ADDISON AV	Celtis australis	Large Tree Routine Prune	10/18/2010
	2	EMERSON ST	Liquidambar styraciflua	Large Tree Routine Prune	6/2/2010
	        
Naive Parsing

Naive parsing of the above data will assume a comma separator and thus results in a single table T with a single column and six rows. The column has the annotations shown in the following table:

idcore annotations
tablenumbersource numbercellstitles
C1T11C1.1, C2.1, C3.1, C4.1, C5.1# publisher City of Palo Alto

The rows have the annotations shown in the following table:

idcore annotations
tablenumbersource numbercells
R1T12C1.1
R2T23C2.1
R3T34C3.1
R4T45C4.1
R5T56C5.1
R6T67C6.1

The cells have the annotations shown in the following table (note that the values of all the cells in the table are strings, denoted by the double quotes in the table below):

idcore annotations
tablecolumnrowstring valuevalue
C1.1TC1R1"# updated 12/31/2010""# updated 12/31/2010"
C1.1TC1R1"#name GID on_street species trim_cycle inventory_date""#name GID on_street species trim_cycle inventory_date"
C2.1TC1R2"#datatype string string string string date:M/D/YYYY""#datatype string string string string date:M/D/YYYY"
C3.1TC1R3" GID On Street Species Trim Cycle Inventory Date"" GID On Street Species Trim Cycle Inventory Date"
C4.1TC1R4" 1 ADDISON AV Celtis australis Large Tree Routine Prune 10/18/2010"" 1 ADDISON AV Celtis australis Large Tree Routine Prune 10/18/2010"
C5.1TC1R5" 2 EMERSON ST Liquidambar styraciflua Large Tree Routine Prune 6/2/2010"" 2 EMERSON ST Liquidambar styraciflua Large Tree Routine Prune 6/2/2010"
Parsing with Flags

The consumer of the data may use the flags described above to create a more useful set of data from this file. Specifically, they could set:

Setting these is done in an implementation-defined way. It could be done, for example, by sniffing the contents of the file itself, through command-line options, or by embedding a dialect description into a metadata file associated with the tabular data, which would look like:

{
  "delimiter": "\t",
  "skipRows": 4,
  "skipColumns": 1,
  "commentPrefix": "#"
}
	          

With these flags in operation, parsing this file results in an annotated tabular data model of a single table T with five columns and two rows which is largely the same as that created from the original simple example described in . There are three differences.

First, because the four skipped rows began with the comment prefix, the table itself now has four rdfs:comment annotations, with the values:

  1. publisher City of Palo Alto
  2. updated 12/31/2010
  3. name GID on_street species trim_cycle inventory_date
  4. datatype string string string string date:M/D/YYYY

Second, because the first column has been skipped, the source number of each of the columns is offset by one from the number of each column:

idcore annotations
tablenumbersource numbercellstitles
C1T12C1.1, C2.1GID
C2T23C1.2, C2.2On Street
C3T34C1.3, C2.3Species
C4T45C1.4, C2.4Trim Cycle
C5T56C1.5, C2.5Inventory Date

Finally, because four additional rows have been skipped, the source number of each of the rows is offset by five from the row number (the four skipped rows plus the single header row):

idcore annotations
tablenumbersource numbercells
R1T16C1.1, C1.2, C1.3, C1.4, C1.5
R2T27C2.1, C2.2, C2.3, C2.4, C2.5
Recognizing Tabular Data Formats

The conventions used in this data (invented for the purpose of this example) are in fact intended to create an annotated tabular data model which includes named annotations on the table itself, on the columns, and on the cells. The creator of these conventions could create a specification for this particular tabular data syntax and register a media type for it. The specification would include statements like:

  • A tab delimiter is always used.
  • The first column is always ignored.
  • When the first column of a row has the value "#", the second column is the name of an annotation on the table and the values of the remaining columns are concatenated to create the value of that annotation.
  • When the first column of a row has the value #name, the remaining cells in the row provide a name annotation for each column in the table.
  • When the first column of a row has the value #datatype, the remaining cells in the row provide datatype/format annotations for the cells within the relevant column, and these are interpreted to create the value for each cell in that column.
  • The first row where the first column is empty is a row of headers; these provide title annotations on the columns in the table.
  • The remaining rows make up the data of the table.

Parsers that recognized the format could then build a more sophisticated annotated tabular data model using only the embedded information in the tabular data file. They would extract embedded metadata looking like:

{
  "@context": "http://www.w3.org/ns/csvw",
  "url": "tree-ops.csv",
  "dc:publisher": "City of Palo Alto",
  "dc:updated": "12/31/2010",
  "tableSchema": {
    "columns": [{
      "name": "GID",
      "titles": "GID",
      "datatype": "string",
    }, {
      "name": "on_street",
      "titles": "On Street",
      "datatype": "string"
    }, {
      "name": "species",
      "titles": "Species",
      "datatype": "string"
    }, {
      "name": "trim_cycle",
      "titles": "Trim Cycle",
      "datatype": "string"
    }, {
      "name": "inventory_date",
      "titles": "Inventory Date",
      "datatype": {
        "base": "date", 
        "format": "M/d/yyyy"
      }
    }]
  }
}
            

As before, the result would be a single table T with five columns and two rows. The table itself would have two annotations:

dc:publisher
{"@value": "City of Palo Alto"}
dc:updated
{"@value": "12/31/2010"}

The columns have the annotations shown in the following table:

idcore annotations
tablenumbersource numbercellsnametitles
C1T12C1.1, C2.1GIDGID
C2T23C1.2, C2.2on_streetOn Street
C3T34C1.3, C2.3speciesSpecies
C4T45C1.4, C2.4trim_cycleTrim Cycle
C5T56C1.5, C2.5inventory_dateInventory Date

The rows have the annotations shown in the following table, exactly as in previous examples:

idcore annotations
tablenumbersource numbercells
R1T16C1.1, C1.2, C1.3, C1.4, C1.5
R2T27C2.1, C2.2, C2.3, C2.4, C2.5

The cells have the annotations shown in the following table. Because of the way the particular tabular data format has been specified, these include additional annotations but also, for the Inventory Date column (cells C1.5 and C2.5), have a value that is a parsed date rather than an unparsed string.

idcore annotations
tablecolumnrowstring valuevalue
C1.1TC1R1"1""1"
C1.2TC2R1"ADDISON AV""ADDISON AV"
C1.3TC3R1"Celtis australis""Celtis australis"
C1.4TC4R1"Large Tree Routine Prune""Large Tree Routine Prune"
C1.5TC5R1"10/18/2010"2010-10-18
C2.1TC1R2"2""2"
C2.2TC2R2"EMERSON ST""EMERSON ST"
C2.3TC3R2"Liquidambar styraciflua""Liquidambar styraciflua"
C2.4TC4R2"Large Tree Routine Prune""Large Tree Routine Prune"
C2.5TC5R2"6/2/2010"2010-06-02

Parsing Multiple Header Lines

The following example shows a CSV file with multiple header lines:

Who,What,,Where,
Organization,Sector,Subsector,Department,Municipality
#org,#sector,#subsector,#adm1,#adm2
UNICEF,Education,Teacher training,Chocó,Quidbó
UNICEF,Education,Teacher training,Chocó,Bojayá
          

Here, the first line contains some grouping titles in the first line, which are not particularly helpful. The lines following those contain useful titles for the columns. Thus the appropriate configuration for a dialect description is:

{
  "skipRows": 1,
  "headerRowCount": 2
}
          

With this configuration, the table model contains five columns, each of which have two titles, summarized in the following table:

idcore annotations
tablenumbersource numbercellstitles
C1T11C1.1, C2.1Organization, #org
C2T22C1.2, C2.2Sector, #sector
C3T33C1.3, C2.3Subsector, #subsector
C4T44C1.4, C2.4Department, #adm1
C5T55C1.5, C2.5Municipality, #adm2

As metadata, this would look like:

{
  "tableSchema": {
    "columns": [
      { "titles": ["Organization", "#org"] }, 
      { "titles": ["Sector", "#sector"] }, 
      { "titles": ["Subsector", "#subsector"] }, 
      { "titles": ["Department", "#adm1"] }, 
      { "titles": ["Municipality", "#adm2"] }, 
    ]
  }
}
          

A separate metadata file could contain just the second of each of these titles, for example:

{
  "tableSchema": {
    "columns": [
      { "name": "org", "titles": #org" }, 
      { "name": "sector", "titles": #sector" }, 
      { "name": "subsector", "titles": #subsector" }, 
      { "name": "adm1", "titles": #adm1" }, 
      { "name": "adm2", "titles": #adm2" }, 
    ]
  }
}
          

This enables people from multiple jurisdictions to use the same tabular data structures without having to use exactly the same titles within their documents.

IANA Considerations

This section has not yet been submitted to IANA for review, approval, and registration.

/.well-known/csvm

URI suffix:
csvm
Change controller:
W3C
Specification document(s):
This document,

Existing Standards

This appendix outlines various ways in which CSV is defined.

RFC 4180

[[!RFC4180]] defines CSV with the following ABNF grammar:

file = [header CRLF] record *(CRLF record) [CRLF]
header = name *(COMMA name)
record = field *(COMMA field)
name = field
field = (escaped / non-escaped)
escaped = DQUOTE *(TEXTDATA / COMMA / CR / LF / 2DQUOTE) DQUOTE
non-escaped = *TEXTDATA
COMMA = %x2C
CR = %x0D
DQUOTE =  %x22
LF = %x0A
CRLF = CR LF
TEXTDATA =  %x20-21 / %x23-2B / %x2D-7E
        

Of particular note here are:

Excel

Excel is a common tool for both creating and reading CSV documents, and therefore the CSV that it produces is a de facto standard.

The following describes the behavior of Microsoft Excel for Mac 2011 with an English locale. Further testing is needed to see the behavior of Excel in other situations.

Saved CSV

Excel generates CSV files encoded using Windows-1252 with LF line endings. Characters that cannot be represented within Windows-1252 are replaced by underscores. Only those cells that need escaping (e.g. because they contain commas or double quotes) are escaped, and double quotes are escaped with two double quotes.

Dates and numbers are formatted as displayed, which means that formatting can lead to information being lost or becoming inconsistent.

Opened CSV

When opening CSV files, Excel interprets CSV files saved in UTF-8 as being encoded as Windows-1252 (whether or not a BOM is present). It correctly deals with double quoted cells, except that it converts line breaks within cells into spaces. It understands CRLF as a line break. It detects dates (formatted as YYYY-MM-DD) and formats them in the default date formatting for files.

Imported CSV

Excel provides more control when importing CSV files into Excel. However, it does not properly understand UTF-8 (with or without BOM). It does however properly understand UTF-16 and can read non-ASCII characters from a UTF-16-encoded file.

A particular quirk in the importing of CSV is that if a cell contains a line break, the final double quote that escapes the cell will be included within it.

Copied Tabular Data

When tabular data is copied from Excel, it is copied in a tab-delimited format, with LF line breaks.

Google Spreadsheets

Downloading CSV

Downloaded CSV files are encoded in UTF-8, without a BOM, and with LF line endings. Dates and numbers are formatted as they appear within the spreadsheet.

Importing CSV

CSV files can be imported as UTF-8 (with or without BOM). CRLF line endings are correctly recognized. Dates are reformatted to the default date format on load.

CSV Files in a Tabular Data Package

Tabular Data Packages place the following restrictions on CSV files:

As a starting point, CSV files included in a Tabular Data Package package must conform to the RFC for CSV (4180 - Common Format and MIME Type for Comma-Separated Values (CSV) Files). In addition:

  • File names MUST end with .csv.

  • Files MUST be encoded as UTF-8.

  • Files MUST have a single header row. This row MUST be the first row in the file.

    • Terminology: each column in the CSV file is termed a field and its name is the string in that column in the header row.

    • The name MUST be unique amongst fields, MUST contain at least one character, and MUST conform to the character restrictions defined for the name property.

  • Rows in the file MUST NOT contain more fields than are in the header row (though they may contain less).

  • Each file MUST have an entry in the tables array in the datapackage.json file.

  • The resource metadata MUST include a tableSchema attribute whose value MUST be a valid schema description.

  • All fields in the CSV files MUST be described in the schema description.

CSV files generated by different applications often vary in their syntax, e.g. use of quoting characters, delimiters, etc. To encourage conformance, CSV files in a CSV files in a Tabular Data Package SHOULD:

  • Use "," as field delimiters.
  • Use CRLF (U+000D U+000A) or LF (U+000A) as line terminators.

If a CSV file does not follow these rules then its specific CSV dialect MUST be documented. The resource hash for the resource in the datapackage.json descriptor MUST:

Applications processing the CSV file SHOULD read use the dialect of the CSV file to guide parsing.

To replicate the findings above, test files which include non-ASCII characters, double quotes, and line breaks within cells are:

Acknowledgements

Changes from previous drafts

Changes since the working draft of 16 April 2015

Changes since the working draft of 08 January 2015

The document has undergone substantial changes since the last working draft. Below are some of the changes made: