<script src="path to js file" type="module"></script> doesn't seem to do anything...

How should the code be structured and separated into different modules?

Considerations:

• keep them context-specific (ex map API endpoint to module)
• optimize for tree-shaking

The issue

"CSV" files created on windows operating systems, will contain a number of different separators and delimiters, depending on the regional settings of the host. This also applies to certain Linux distributions (looking at you ubuntu).
In the current implementation only the comma separator(,) and double-quote delimiter(") are supported, as stated in the RFC.

requested feature

It should be possible to define custom separator and delimiter values, via the option object. These values should be optional and default to comma and double-quote.

Currently, the parser isn't very forgiving when it comes to excess whitespace

Changes

• ignore leading whitespace on unquoted values
• ignore trailing whitespace on unquoted values
• ignore leading whitespace on quoted values
• ignore trailing whitespace on quoted values
• full test coverage for the new functionality

Readme mentions A CommonJS bundle is included for backward compatible with node <= 13.2 but package does not even install on node <13.2 yarn add gives error of node version.

So how are we supposed to use this package on node < 13.2

Thanks.

Hi - thanks for sharing your great CSV handling code.
One thing - I think the spec allows for not having a newline at the end of the file:
2. The last record in the file may or may not have an ending line
break. For example:
aaa,bbb,ccc CRLF
zzz,yyy,xxx
but I don't think your code does yet. Hopefully easily fixed.

By default, CSV is stringly-typed. Ie everything is a string. Since stringly-typed values aren't very useful in a software context, this feature exists to add automatic type inference to the parser.

Changes

• add a private inferType() function
• add options.typed to CSV.parse() [default: false]
• hook inferType() into the parser

Specifics:

The type inference castToScalar() function from the jquery-csv package can be reused here.

castToScalar: function (value, state) {
        var hasDot = /\./;
        if (isNaN(value)) {
          return value;
        } else {
          if (hasDot.test(value)) {
            return parseFloat(value);
          } else {
            var integer = parseInt(value);
            if (isNaN(integer)) {
              return null;
            } else {
              return integer;
            }
          }
        }
      }

For readability, it's probably a good idea to refactor this as a switch statement.

Source: $.castToScalar()

The replacer allows users to hook into the formatter with their own function.

It should expose the

• value being formatted
• row of the value being formatted
• column of the value being formatted

Changes

• add row/col state tracking to CSV.stringify
• hook the optional function call into the formatter

Check the inputs to ensure they're the correct types. Throw if not

Checklist

• CSV.parse inputs
• CSV.stringify inputs

The reviver parameter allows users hook a custom function into the parser.

It should provide the following

• the value being parsed
• the row of the value being parsed
• the column of the value being parsed

The row/col states tracking is already included in the parser.

Changes

• hook the reviver into the parser implementation

README.md documents order of reviver arguments as

reviver(value, col, row)

but implementation seems to pass value, row, col, at least for CSV.parse.

Also, documentation doesn't say whether row and col are 1-based or 0-based. Seems to be 1.

What major endpoints should this expose?

Considerations:

• simple is 👍
• familiar is 👍
• intuitive is 👍

Hi all! This is good work and I'm happy to find a pure-ES CSV parser.

I have some data I'm trying to parse, but it throws errors. Here is an example:

export var town06_csv = `
x,y,z,pitch,yaw,roll,formatted
586.8056030273438, -10.063207626342773, 0.29999998211860657, 0.0, -179.58056640625, 0.0, "[586.806, -10.063, 0.300, 0.000, -179.581, 0.000]"
584.8312377929688, -13.57775592803955, 0.29999998211860657, 0.0, -179.58056640625, 0.0, "[584.831, -13.578, 0.300, 0.000, -179.581, 0.000]"
`

Note the last entry is "[586.806, -10.063, 0.300, 0.000, -179.581, 0.000]", i.e. a single entry enclosed in double quotes. I import this CSV string and the parser, and run parse like this:

<script type="module">
  import { town06_csv } from './raw_spawn_data.js';
  import { parse } from './escsv.js'
  var town06_spawnpoints = parse(town06_csv);
</script>

But we get an error Uncaught Error: CSVError: Illegal state [row:3, col:7]! Per RFC4180 ( https://www.rfc-editor.org/rfc/rfc4180#section-2 ), this should be allowed, because the field is enclosed by a double quote. As far as I can tell, there is no error in my CSV, and it parses correctly in other parsers, e.g. in OpenOffice:

Specifically, it throws the error on line 67 (case 2, un-delimited input, defaults to state 4.)

Hi, great library!

Found an issue when serializing a field with more than one double quote.
For instance a value of John Dwayne "The Rock" Johnson is serialized as "Dwayne ""The Rock" Johnson".
When trying to parse that with CSV.parse it throws CSVError: Illegal state.

Code snippet:
const CSV = require('@vanillaes/csv'); const data = [ [ "header1", "header2", "header3" ], [ "aaa", "bbb", 'Dwayne "The Rock" Johnson' ], [ "zzz", "yyy", "xxx" ] ]; const stringified = CSV.stringify(data); const rows = CSV.parse(stringified);

Throws:
Error: CSVError: Illegal state [row:2, col:3]

When importing csv-es into a TypeScript project via import CSV from 'csv-es the following errors appear at compile time:

ERROR in node_modules/csv-es/index.d.ts(28,14): error TS2303: Circular definition of import alias 'parse'. node_modules/csv-es/index.d.ts(29,14): error TS2303: Circular definition of import alias 'stringify'.

What sort of repo maintenance/tooling should this project include?

Testing

tape.js + tap-spec

Tape was a great choice for jquery-csv.

It:

• is small
• has a minimal API
• outputs in TAP format, which can be piped into a ton of different transforms
• compatible with ES Modules (with workarounds)

Tape-es for tape w/ ESM support

Should code-coverage be included?

No

Linting

SemiStandard

Continuous Integration

TravisCI works well for checking if the build is broken but this package should include auto-publish to NPM on tag push.

Circle-CI is a pretty good alternative.

GitHub Actions

Minification

None

HOLD: This feature is on hold, I'm not 100% convinced it's a necessary addition

If set, the options.headers should exclude the first row of values if it's set to true.

Changes

• include the headers option
• hook it into the formatter

Hi, it seems there is a bug where single column csv content that does not end with a line break gets the last "cell" value deleted when parsed. I have created a PR where I updated the tests to expose the issue, but do not have enough knowledge to fix the issue. I have tried updating the code at https://github.com/vanillaes/csv/blob/main/index.js#L104-L107 to honour flushing of the last value using if (ctx.value !== '') instead but that seems to break some other tests.

Any ideas on how to fix this would be appreciated.

Cheers

@coltonehrman Add yourself to the contributors list and -- if you want -- the license.

BTW, package.json only allows one author/project so I added myself as a contributor instead.

HOLD: This feature is on hold, I'm not 100% convinced it's a necessary addition

If set options.headers should take the headers into account. If type inference is set, the headers row should be skipped.