[FEATURE] Http Request Transformation in UDF

[Ttommeke]

This PR makes it possible to execute a Http request from a node in the cluster rather than collecting the data on the driver and executing the request there.

Description

In this PR a UDF was created that can handle http requests. The implementation is based on the socket and ssl library because the requests library is not able to be pickeled. The UDF can handle multiple scenarios GET, POST, PUT, DELETE, ... With authorization and a body to be send.

Proper documentation still needs to be written. Waiting for any remarks before doing so.

download file transformation

Motivation and Context

This is a suggestion to make the download file transformation proces better. Looking for feedback on the usecase.

How Has This Been Tested?

A test is written.

Screenshots (if appropriate):

Types of changes

• Bug fix (non-breaking change which fixes an issue)
• New feature (non-breaking change which adds functionality)
• Breaking change (fix or feature that would cause existing functionality to change)

Checklist:

• My code follows the code style of this project.
• My change requires a change to the documentation.
• I have updated the documentation accordingly.
• I have read the CONTRIBUTING document.
• I have added tests to cover my changes.
• All new and existing tests passed.

[dannymeijer]

Overall, good job :) - left you some comments (nothing too major)

[dannymeijer]

let's make these constants, meaning:

• you don't have to start with an underscore (no need to mark them private)
• make them all caps: "CHUNK_SIZE_COLUMN" for example
• you can remove the "HttpRequestTransformation" part from them (for brevity)

[Ttommeke]

Thanks!

Deleted the statements anyway since we didn't use them.

[dannymeijer]

make this a SecretStr - we don't want tokens to be printed unnecessarily

[dannymeijer]

from koheesio.models import SecretStr

[Ttommeke]

Thanks!

[dannymeijer]

right type would be Optional[str] rather than Union[str, None] (same for the other Fields where you have it like this)

[Ttommeke]

check!

[dannymeijer]

let's move the udf to the top of the file, makes it a bit more readable

[Ttommeke]

Okay!

[dannymeijer]

could we do this with regex instead? Seems a bit cleaner than this imho.

[dannymeijer]

for example:

import re

import pytest

url_parts = re.compile(
    # protocol is a sequence of letters, digits, and characters followed by ://
    r'^(?P<protocol>[^:]+)://'
    # base_url is a sequence of characters until encountering a `/` or a `:` followed by 1 to 5 digits
    r'(?P<base_url>\S+?(?=:\d{1,5}/|/|$))'
    # port is an optional sequence of digits between 0 and 65535 preceded by a colon
    r'(?::(?P<port>\d{1,5}))?'
    # rest is the rest of the url
    r'(?P<rest>\/.*)?$'
)

test_data = [
    # url, expected_protocol, expected_base_url, expected_port, expected_rest
    ("https://example.com:443/some/random/url", "https", "example.com", "443", "/some/random/url"),
    ("http://something.a.bit.more.complicated:9999/foo/bar", "http", "something.a.bit.more.complicated", "9999", "/foo/bar"),
    ("https://no-port-example.ext/42/index.jsp?foo=bar&baz=bla", "https", "no-port-example.ext", None, "/42/index.jsp?foo=bar&baz=bla"),
    ("ftp://ftp.example.com/resource.txt", "ftp", "ftp.example.com", None, "/resource.txt"),
    ("http://localhost:8080/test", "http", "localhost", "8080", "/test"),
    ("https://sub.domain.example.com/path/to/resource?query=string&another=value", "https", "sub.domain.example.com", None, "/path/to/resource?query=string&another=value"),
    ("http://192.168.1.1:8080/admin", "http", "192.168.1.1", "8080", "/admin"),
    ("https://user:[email protected]:8443/path?query=param#fragment", "https", "user:[email protected]", "8443", "/path?query=param#fragment"),
    ("http://example.org", "http", "example.org", None, None),
    ("https://example.net/path/to/resource.html", "https", "example.net", None, "/path/to/resource.html"),
    ("http://example.com:80/path/to/resource?query=param", "http", "example.com", "80", "/path/to/resource?query=param"),
    ("custom_protocol://base_url:foo/rest", "custom_protocol", "base_url:foo", None, "/rest"),
    ("WEBDAV://example.com:8080/path/to/resource", "WEBDAV", "example.com", "8080", "/path/to/resource"),
]

@pytest.mark.parametrize("url, expected_protocol, expected_base_url, expected_port, expected_rest", test_data)
def test_url_parts(url, expected_protocol, expected_base_url, expected_port, expected_rest):
    match = url_parts.match(url)

    assert match is not None, f"Failed to match {url}"
    assert match.group("protocol") == expected_protocol, f"Failed to match protocol: {match.groupdict()}"
    assert match.group("port") == expected_port, f"Failed to match port: {match.groupdict()}"
    assert match.group("base_url") == expected_base_url, f"Failed to match base_url: {match.groupdict()}"
    assert match.group("rest") == expected_rest, f"Failed to match rest: : {match.groupdict()}"

[Ttommeke]

I would argue the other way around but Regex it is.

[dannymeijer]

You might be right.. (I am biased to regex, granted). Please run the above tests through your code and see if they pass as expected; if they do, I don't mind not using regex.

[dannymeijer]

response += chunk

[dannymeijer]

I'm pretty sure I saved the expected data in a file already in the data folder of the tests right? Either way, I would like to take this hardcoding out of the code and refer to a file instead.

[Ttommeke]

Not 100% sure if it doesn't happen somewhere else, but the test_downloading_files is only checking the dataframe and if the files exist. not the contents of the downloaded file.

[Ttommeke]

I'm gonna put the content in test files btw. ;)