Some validation errors are not being identified

Mikanebu commented 6 years ago

Reported here: datahq/datahub-qa#68

We want to improve showcase page, so it should have error messages as per readme of the dp, eg, when pushing following datapackages some errors are not catched in the frontend:

https://github.com/frictionlessdata/test-data/tree/master/packages/types-formats-and-constraints

string - The value “https:/domain.com” in column “uri” is not type “string” and format “uri”
datetime - The value “2018/01/02T00:00:00” in column “any” is not type “datetime” and format “any”
time - The value “00:01” in column “any” is not type “time” and format “any”

Acceptance criteria

the dataset page should have error messages on showcase page as per readme of the dp:

[x] string - The value “https:/domain.com” in column “uri” is not type “string” and format “uri”
- WONTFIX: https://github.com/datahq/frontend/issues/191#issuecomment-370723362
[x] datetime - The value “2018/01/02T00:00:00” in column “any” is not type “datetime” and format “any”
- WONTFIX: https://github.com/datahq/frontend/issues/191#issuecomment-370764258
[x] time - The value “00:01” in column “any” is not type “time” and format “any”
- WONTFIX: '00:01' is not time? :face_with_head_bandage: Really?!! Who's that guy, who created the corresponding test?

Tasks

[ ] add tests for tableschema-py
[ ] fix tableschema-py

Analysis

data validate uses tableschema-js and found all errors
server users tableschema-py and NOT found some errors

How to reproduce

data validate and remember errors
data push and see errors on the dataset page

string: $ data validate test-data/packages/types-formats-and-constraints/string/ $ data push test-data/packages/types-formats-and-constraints/string/
datetime: $ data validate test-data/packages/types-formats-and-constraints/datetime/ $ data push test-data/packages/types-formats-and-constraints/datetime/
time: $ data validate test-data/packages/types-formats-and-constraints/time/ $ data push test-data/packages/types-formats-and-constraints/time/

AcckiyGerman commented 6 years ago

WONTFIX: String validator accept `http:/datahub.io` uri

The difference between JS and Python versions of tableschema about validating URI is:

tableschema-js uses "validator": "^9.2.0" from npm
tableschema-py uses https://pypi.python.org/pypi/rfc3986

AND

http:/datahub.io is NOT VALID for npm validator
http:/datahub.io is VALID for pypi rfc3986

When I goes deeper I realized that there is an ANCIENT & GREAT HOLY WAR called: Accept or not Accept an URI with less or more than two // separators:

https://daniel.haxx.se/blog/2016/05/11/my-url-isnt-your-url/ The WHATWG spec says it has to be one slash and that a parser must accept an indefinite amount of slashes. “http:/example.com” and “http:////////////////////////////////////example.com” are both equally fine. RFC 3986 and many others would disagree.

Also, most of browsers accept http:/datahub.io uri - I just have checked.

So the pypi implementation of rfc3986 DOES accepts this kind of invalid URI; And I doubt that @roll will accept any PR where we will implement URI validator by ourselves, just for fixing this particular edge case.

Also, the URI validating REGEX is so f**king complex:

/
# protocol user host-ip port path path path querystring fragment
^
#protocol
(?:(?<scheme>[a-zA-Z][a-zA-Z\d+-.]*):)?
(?:
  (?:
    (?:
        \/\/
        (?:
            #userinfo
            (?:((?:[a-zA-Z\d\-._~\!$&'()*+,;=%]*)(?::(?:[a-zA-Z\d\-._~\!$&'()*+,;=:%]*))?)@)?
            #host-ip
            ((?:[a-zA-Z\d-.%]+)|(?:\d{1,3}\.\d{1,3}\.\d{1,3}\.\d{1,3})|(?:\[(?:[a-fA-F\d.:]+)\]))?
            #port
            (?::(\d*))?
        )
    )
    #slash-path
    (
        (?:\/[a-zA-Z\d\-._~\!$&'()*+,;=:@%]*)*
    )
  )
 #slash-path
 |(\/(?:(?:[a-zA-Z\d\-._~\!$&'()*+,;=:@%]+(?:\/[a-zA-Z\d\-._~\!$&'()*+,;=:@%]*)*))?)
 #path
 |([a-zA-Z\d\-._~\!$&'()*+,;=:@%]+(?:\/[a-zA-Z\d\-._~\!$&'()*+,;=:@%]*)*)
)?
#querystring
(?:\?([a-zA-Z\d\-._~\!$&'()*+,;=:@%\/?]*))?
#fragment
(?:\#([a-zA-Z\d\-._~\!$&'()*+,;=:@%\/?]*))?
$
/x

So I'm leaving this URI validator as it is.

AcckiyGerman commented 6 years ago

WONTFIX: `tableschema-py` validator accepts datetime `2018/01/02T00:00:00` format: Any

The same situation here - py and js versions of tableschema use different libs to validate datetime

This 2018/01/02T00:00:00 is a

Valid date due to python lib

>>> from dateutil.parser import parse
>>> parse("2018/01/02T00:00:00")
datetime.datetime(2018, 1, 2, 0, 0)

Invalid date due to JS libs

> moment = require('moment')
> moment("2018/01/02T00:00:00")
Deprecation warning: value provided is not in a recognized RFC2822 or ISO format. moment construction falls back to js Date(), which is not reliable across all browsers and versions.

So my opinion - we should not restrict tableschema-py only because tableschema-js could not recognize some valid dates.

AcckiyGerman commented 6 years ago

FIXED: JS version of tableschema gives some validation errors, while PY version doesn't:

00:01 is not Time
"2018/01/02T00:00:00" is not datetime with format 'Any'
http:/datahub.io is not correct URI (I agree with the last one, but developers of rfc3986 validator lib for Python are not agree)

So, everything is good with our Pipeline.

If we think that 00:01 is valid time format - we should fix tableschema-js (and data validate) instead.
If we think that 00:01 is invalid time format - then we should validate a dataset prior pushing, coz tableschema-py Authors will not accept such a change anyway.

And I'm going to change the corresponding QA test.

datahub-v2 / frontend