fix: IRI to URI conversion (#2304) · RDFLib/rdflib@dfa4054 (original) (raw)

`@@ -522,32 +522,92 @@ def _coalesce(

522

`return default

523

524

525

_RFC3986_SUBDELIMS = "!$&'()*+,;="

526

"""

527


``sub-delims`` production from `RFC 3986, section 2.2

528

`` +

https://www.rfc-editor.org/rfc/rfc3986.html#section-2.2`_.

529

"""

530

+

531

_RFC3986_PCHAR_NU = "%" + _RFC3986_SUBDELIMS + ":@"

532

"""

533


The non-unreserved characters in the ``pchar`` production from RFC 3986.

534

"""

535

+

536

_QUERY_SAFE_CHARS = _RFC3986_PCHAR_NU + "/?"

537

"""

538

The non-unreserved characters that are safe to use in in the query and fragment

539

components.

540

+

541

.. code-block::

542

+

543

pchar = unreserved / pct-encoded / sub-delims / ":" / "@" query

544

= *( pchar / "/" / "?" ) fragment = *( pchar / "/" / "?" )

545

"""

546

+

547

_USERNAME_SAFE_CHARS = _RFC3986_SUBDELIMS + "%"

548

"""

549

The non-unreserved characters that are safe to use in the username and password

550

components.

551

+

552

.. code-block::

553

+

554

userinfo = *( unreserved / pct-encoded / sub-delims / ":" )

555

+

556

":" is excluded as this is only used for the username and password components,

557

and they are treated separately.

558

"""

559

+

560

_PATH_SAFE_CHARS = _RFC3986_PCHAR_NU + "/"

561

"""

562

The non-unreserved characters that are safe to use in the path component.

563

+

564

+

565

This is based on various path-related productions from RFC 3986.

566

"""

567

+

568

+

525

569

`def _iri2uri(iri: str) -> str:

526

570

`"""

527

Convert an IRI to a URI (Python 3).

528

https://stackoverflow.com/a/42309027

529

https://stackoverflow.com/a/40654295

530

netloc should be encoded using IDNA;

531

non-ascii URL path should be encoded to UTF-8 and then percent-escaped;

532

non-ascii query parameters should be encoded to the encoding of a page

533

URL was extracted from (or to the encoding server uses), then

534

percent-escaped.

571

Prior art:

572

+

573

`` +

iri_to_uri from Werkzeug <https://github.com/pallets/werkzeug/blob/92c6380248c7272ee668e1f8bbd80447027ccce2/src/werkzeug/urls.py#L926-L931>_

574

+

535

575

` >>> _iri2uri("https://dbpedia.org/resource/Almería")

536

576

` 'https://dbpedia.org/resource/Almer%C3%ADa'

537

577

` """

578

https://datatracker.ietf.org/doc/html/rfc3986

538

579

`# https://datatracker.ietf.org/doc/html/rfc3305

539

580

540

(scheme, netloc, path, query, fragment) = urlsplit(iri)

581

parts = urlsplit(iri)

582

(scheme, netloc, path, query, fragment) = parts

541

583

542

Just support http/https, otherwise return the iri unmolested

584

Just support http/https, otherwise return the iri unaltered

543

585

`if scheme not in ["http", "https"]:

544

586

`return iri

545

587

546

scheme = quote(scheme)

547

netloc = netloc.encode("idna").decode("utf-8")

548

path = quote(path)

549

query = quote(query)

550

fragment = quote(fragment)

588

path = quote(path, safe=_PATH_SAFE_CHARS)

589

query = quote(query, safe=_QUERY_SAFE_CHARS)

590

fragment = quote(fragment, safe=_QUERY_SAFE_CHARS)

591

+

592

if parts.hostname:

593

netloc = parts.hostname.encode("idna").decode("ascii")

594

else:

595

netloc = ""

596

+

597

if ":" in netloc:

598

Quote IPv6 addresses

599

netloc = f"[{netloc}]"

600

+

601

if parts.port:

602

netloc = f"{netloc}:{parts.port}"

603

+

604

if parts.username:

605

auth = quote(parts.username, safe=_USERNAME_SAFE_CHARS)

606

if parts.password:

607

pass_quoted = quote(parts.password, safe=_USERNAME_SAFE_CHARS)

608

auth = f"{auth}:{pass_quoted}"

609

netloc = f"{auth}@{netloc}"

610

+

551

611

`uri = urlunsplit((scheme, netloc, path, query, fragment))

552

612

553

613

`if iri.endswith("#") and not uri.endswith("#"):