help: improve view and list commands

Ultimately a simple link list does not do all you want, so now the
history maintains a list of visited links that are useful to display,
but also a backlink list to let the user move to the previous pages.

BOARD.txt

@@ -4,7 +4,7 @@ TODO                                    DONE
                                         links
                                         redirections
                                         web links
-                                        history (back/forward)
+                                        history (back)
                                         simple caching
                                         simple text files
                                         encodings
@@ -14,20 +14,26 @@ TODO                                    DONE
                                         configuration
                                         help page
                                         TOFU
-open last download
+                                        view history
+                                        open last download
+                                        media files
 home page
-media files
-view history
 identity management
 --------------------------------------------------------------------------------
 BACKLOG
 click on links to open them
 download to disk, not in memory
+download in the background
+download view instead of last download
 does encoding really work? cf. egsam
 margins / centering
 pre blocks folding
 buffers (tabs)
+a11y? tts?
 handle soft-hyphens on wrapping
 bug: combining chars reduce lengths
 non shit command-line
 response code 11 (if still there)
+gopher?
+save history
+history (forward) (useful?)

README.md

@@ -57,7 +57,7 @@ It also provide these features:
 
 - History
 - Caching
-- Bookmarks: it's just a text file with bindings.
+- Bookmarks (it's just a text file with bindings)
 - Downloads
 
 Check out [this board](BOARD.txt) for what's done and coming next.
@@ -74,11 +74,14 @@ you want, just restart Bebop to take changes into account.
 Here are the available options:
 
 | Key                        | Type         | Default        | Description                           |
-|-------------------|-------------|----------|---------------------------------------|
+|----------------------------|--------------|----------------|---------------------------------------|
 | `connect_timeout`          | int          | 10             | Seconds before connection times out.  |
 | `text_width`               | int          | 80             | Rendered line length.                 |
 | `source_editor`            | string list  | `["vi"]`       | Command to use for editing sources.   |
 | `command_editor`           | string list  | `["vi"]`       | Command to use for editing CLI input. |
+| `history_limit`            | int          | 1000           | Maximum entries in history.           |
+| `external_commands`        | (see note 2) | {}             | Commands to open various files.       |
+| `external_command_default` | string list  | `["xdg-open"]` | Default command to open files.        |
 
 Note: for the "command" parameters such as `source_editor` and `command_editor`,
 a string list is used to separate the different program arguments, e.g. if you
@@ -86,6 +89,12 @@ wish to use `vim -c 'startinsert'`, you should write the list `["vim", "-c",
 "startinsert"]`. In both case, a temporary or regular file name will be appended
 to this command when run.
 
+2: the `external_commands` dict maps MIME types to commands just as above. For
+example, if you want to open video files with VLC and audio files in Clementine,
+you can use the following dict: `{"audio": ["clementine"], "video", ["vlc"]}`.
+For now only "main" MIME types are supported, i.e. you cannot specify precise
+types like "audio/flac", just "audio".
+
 
 
 FAQ

bebop/browser/browser.py

@@ -4,8 +4,11 @@ import curses
 import curses.ascii
 import curses.textpad
 import os
+import subprocess
 import tempfile
 from math import inf
+from pathlib import Path
+from typing import Optional, Tuple
 
 from bebop.bookmarks import (
     get_bookmarks_path, get_bookmarks_document, save_bookmark
@@ -16,9 +19,11 @@ from bebop.external import open_external_program
 from bebop.help import HELP_PAGE
 from bebop.history import History
 from bebop.links import Links
+from bebop.mime import MimeType
 from bebop.mouse import ButtonState
 from bebop.navigation import (
-    get_parent_url, get_root_url, join_url, parse_url, sanitize_url)
+    get_parent_url, get_root_url, join_url, parse_url, unparse_url
+)
 from bebop.page import Page
 from bebop.page_pad import PagePad
 
@@ -43,6 +48,7 @@ class Browser:
       values are dicts as well: the "open" key maps to a callable to use when
       the page is accessed, and the optional "source" key maps to callable
       returning the page source path.
+    - last_download: tuple of MimeType and path, or None.
     """
 
     def __init__(self, config, cert_stash):
@@ -55,9 +61,10 @@ class Browser:
         self.command_line = None
         self.running = True
         self.status_data = ("", 0, 0)
-        self.history = History()
+        self.history = History(self.config["history_limit"])
         self.cache = {}
         self.special_pages = self.setup_special_pages()
+        self.last_download: Optional[Tuple[MimeType, Path]] = None
         self._current_url = ""
 
     @property
@@ -89,6 +96,9 @@ class Browser:
             "help": {
                 "open": self.open_help,
             },
+            "history": {
+                "open": self.open_history,
+            },
         }
 
     def run(self, *args, **kwargs):
@@ -164,6 +174,8 @@ class Browser:
             self.scroll_page_vertically(inf)
         elif char == ord("o"):
             self.quick_command("open")
+        elif char == ord("O"):
+            self.open_last_download()
         elif char == ord("p"):
             self.go_back()
         elif char == ord("u"):
@@ -176,6 +188,8 @@ class Browser:
             self.add_bookmark()
         elif char == ord("e"):
             self.edit_page()
+        elif char == ord("y"):
+            self.open_history()
         elif curses.ascii.isdigit(char):
             self.handle_digit_input(char)
         elif char == curses.KEY_MOUSE:
@@ -197,10 +211,15 @@ class Browser:
                 elif char == ord("l"):
                     self.scroll_page_horizontally(1)
             self.screen.nodelay(False)
-        # else:
-        #     unctrled = curses.unctrl(char)
-        #     if unctrled == b"^T":
-        #         self.set_status("test!")
+        # elif char == ord("@"):
+        #     self.current_url = "bebop:debugzone"
+        #     t = "\n".join("* " + u for u in self.history.urls)
+        #     t += "\n\n" + "\n".join("* " + u for u in self.history.backlist)
+        #     self.load_page(Page.from_text(t))
+        #     # unctrled = curses.unctrl(char)
+        #     # if unctrled == b"^T":
+        #     #     self.set_status("test!")
+        #     pass
 
     @property
     def page_pad_size(self):
@@ -300,38 +319,52 @@ class Browser:
             return
 
         if assume_absolute or not self.current_url:
-            parts = parse_url(url, absolute=True)
-            join = False
+            parts = parse_url(url, absolute=True, default_scheme="gemini")
         else:
             parts = parse_url(url)
-            join = True
 
-        if parts.scheme == "gemini":
+        if parts["scheme"] is None and parts["netloc"] is None:
+            base_url = base_url or self.current_url
+            if base_url:
+                parts = parse_url(join_url(base_url, url))
+            else:
+                self.set_status_error(f"Can't open '{url}'.")
+                return
+
+        # Replace URL passed as parameter by a proper absolute one.
+        url = unparse_url(parts)
+
+        scheme = parts["scheme"] or ""
+        if scheme == "gemini":
             from bebop.browser.gemini import open_gemini_url
-            # If there is no netloc, this is a relative URL.
-            if join or base_url:
-                url = join_url(base_url or self.current_url, url)
-            open_gemini_url(
+            success = open_gemini_url(
                 self,
-                sanitize_url(url),
+                url,
                 redirects=redirects,
-                history=history,
                 use_cache=use_cache
             )
-        elif parts.scheme.startswith("http"):
+            if history and success:
+                self.history.push(url)
+
+        elif scheme.startswith("http"):
             from bebop.browser.web import open_web_url
             open_web_url(self, url)
-        elif parts.scheme == "file":
+
+        elif scheme == "file":
             from bebop.browser.file import open_file
-            open_file(self, parts.path, history=history)
-        elif parts.scheme == "bebop":
-            special_page = self.special_pages.get(parts.netloc)
+            file_url = open_file(self, parts["path"])
+            if history and file_url:
+                self.history.push(file_url)
+
+        elif scheme == "bebop":
+            special_page = self.special_pages.get(parts["path"])
             if special_page:
                 special_page["open"]()
             else:
                 self.set_status_error("Unknown page.")
+
         else:
-            self.set_status_error(f"Protocol {parts.scheme} not supported.")
+            self.set_status_error(f"Protocol '{scheme}' not supported.")
 
     def load_page(self, page: Page):
         """Load Gemtext data as the current page."""
@@ -455,8 +488,12 @@ class Browser:
 
     def go_back(self):
         """Go back in history if possible."""
-        if self.history.has_links():
-            self.open_url(self.history.pop(), history=False)
+        if self.current_url.startswith("bebop:"):
+            previous_url = self.history.get_previous(actual_previous=True)
+        else:
+            previous_url = self.history.get_previous()
+        if previous_url:
+            self.open_url(previous_url, history=False)
 
     def go_to_parent_page(self):
         """Go to the parent URL if possible."""
@@ -468,14 +505,19 @@ class Browser:
         if self.current_url:
             self.open_url(get_root_url(self.current_url))
 
+    def open_internal_page(self, name, gemtext):
+        """Open some content corresponding to a "bebop:" internal URL."""
+        page = Page.from_gemtext(gemtext, self.config["text_width"])
+        self.load_page(page)
+        self.current_url = "bebop:" + name
+
     def open_bookmarks(self):
         """Open bookmarks."""
         content = get_bookmarks_document()
         if content is None:
             self.set_status_error("Failed to open bookmarks.")
             return
-        self.load_page(Page.from_gemtext(content, self.config["text_width"]))
-        self.current_url = "bebop://bookmarks"
+        self.open_internal_page("bookmarks", content)
 
     def add_bookmark(self):
         """Add the current URL as bookmark."""
@@ -502,8 +544,9 @@ class Browser:
         directly from their location on disk.
         """
         delete_source_after = False
-        if self.current_url.startswith("bebop://"):
-            page_name = self.current_url[len("bebop://"):]
+        parts = parse_url(self.current_url)
+        if parts["scheme"] == "bebop":
+            page_name = parts["path"]
             special_pages_functions = self.special_pages.get(page_name)
             if not special_pages_functions:
                 return
@@ -529,10 +572,33 @@ class Browser:
 
     def open_help(self):
         """Show the help page."""
-        self.load_page(Page.from_gemtext(HELP_PAGE, self.config["text_width"]))
-        self.current_url = "bebop://help"
+        self.open_internal_page("help", HELP_PAGE)
 
     def prompt(self, text, keys):
         """Display the text and allow it to type one of the given keys."""
         self.set_status(text)
         return self.command_line.prompt_key(keys)
+
+    def open_history(self):
+        """Show a generated history of visited pages."""
+        self.open_internal_page("history", self.history.to_gemtext())
+
+    def open_last_download(self):
+        """Open the last downloaded file."""
+        if not self.last_download:
+            return
+        mime_type, path = self.last_download
+        command = self.config["external_commands"].get(mime_type.main_type)
+        if not command:
+            command = self.config["external_command_default"]
+        command = command + [str(path)]
+        self.set_status(f"Running '{' '.join(command)}'...")
+        try:
+            subprocess.Popen(
+                command,
+                stdout=subprocess.DEVNULL,
+                stderr=subprocess.DEVNULL,
+                start_new_session=True
+            )
+        except FileNotFoundError as exc:
+            self.set_status_error(f"Failed to run command: {exc}")

bebop/browser/file.py

@@ -4,22 +4,29 @@ from bebop.browser.browser import Browser
 from bebop.page import Page
 
 
-def open_file(browser: Browser, filepath: str, encoding="utf-8", history=True):
+def open_file(browser: Browser, filepath: str, encoding="utf-8"):
     """Open a file and render it.
 
     This should be used only on Gemtext files or at least text files.
     Anything else will produce garbage and may crash the program. In the
     future this should be able to use a different parser according to a MIME
     type or something.
+
+    Arguments:
+    - browser: Browser object making the request.
+    - filepath: a text file path on disk.
+    - encoding: file's encoding.
+
+    Returns:
+    The loaded file URI on success, None otherwise (e.g. file not found).
     """
     try:
         with open(filepath, "rt", encoding=encoding) as f:
             text = f.read()
     except (OSError, ValueError) as exc:
         browser.set_status_error(f"Failed to open file: {exc}")
-        return
+        return None
     browser.load_page(Page.from_text(text))
     file_url = "file://" + filepath
-    if history:
-        browser.history.push(file_url)
     browser.current_url = file_url
+    return file_url

bebop/browser/gemini.py

@@ -14,8 +14,7 @@ from bebop.tofu import trust_fingerprint, untrust_fingerprint, WRONG_FP_ALERT
 MAX_URL_LEN = 1024
 
 
-def open_gemini_url(browser: Browser, url, redirects=0, history=True,
-                    use_cache=True):
+def open_gemini_url(browser: Browser, url, redirects=0, use_cache=True):
     """Open a Gemini URL and set the formatted response as content.
 
     While the specification is not set in stone, every client takes a slightly
@@ -33,12 +32,14 @@ def open_gemini_url(browser: Browser, url, redirects=0, history=True,
       as we're doing TOFU here, we could automatically trust it or let the user
       choose. For simplicity, we always trust it permanently.
 
-    Attributes:
+    Arguments:
     - browser: Browser object making the request.
     - url: a valid URL with Gemini scheme to open.
     - redirects: current amount of redirections done to open the initial URL.
-    - history: if true, save the final URL to history.
     - use_cache: if true, look up if the page is cached before requesting it.
+
+    Returns:
+    True on success, False otherwise.
     """
     if len(url) >= MAX_URL_LEN:
         browser.set_status_error("Request URL too long.")
@@ -48,11 +49,9 @@ def open_gemini_url(browser: Browser, url, redirects=0, history=True,
 
     if use_cache and url in browser.cache:
         browser.load_page(browser.cache[url])
-        if browser.current_url and history:
-            browser.history.push(browser.current_url)
         browser.current_url = url
         browser.set_status(url)
-        return
+        return True
 
     req = Request(url, browser.stash)
     connect_timeout = browser.config["connect_timeout"]
@@ -69,7 +68,7 @@ def open_gemini_url(browser: Browser, url, redirects=0, history=True,
         else:
             error = f"Connection failed ({url})."
         browser.set_status_error(error)
-        return
+        return False
 
     if req.state == Request.STATE_INVALID_CERT:
         pass
@@ -88,13 +87,13 @@ def open_gemini_url(browser: Browser, url, redirects=0, history=True,
     data = req.proceed()
     if not data:
         browser.set_status_error(f"Server did not respond in time ({url}).")
-        return
+        return False
     response = Response.parse(data)
     if not response:
         browser.set_status_error(f"Server response parsing failed ({url}).")
-        return
+        return False
 
-    _handle_response(browser, response, url, redirects, history)
+    return _handle_response(browser, response, url, redirects)
 
 
 def _handle_untrusted_cert(browser: Browser, request: Request):
@@ -118,10 +117,14 @@ def _handle_untrusted_cert(browser: Browser, request: Request):
 
 
 def _handle_response(browser: Browser, response: Response, url: str,
-                     redirects: int, history: bool):
-    """Handle a response from a Gemini server."""
+                     redirects: int):
+    """Handle a response from a Gemini server.
+
+    Returns:
+    True on success, False otherwise.
+    """
     if response.code == 20:
-        _handle_successful_response(browser, response, url, history)
+        return _handle_successful_response(browser, response, url)
     elif response.generic_code == 30 and response.meta:
         browser.open_url(response.meta, base_url=url, redirects=redirects + 1)
     elif response.generic_code in (40, 50):
@@ -132,10 +135,10 @@ def _handle_response(browser: Browser, response: Response, url: str,
     else:
         error = f"Unhandled response code {response.code}"
         browser.set_status_error(error)
+    return False
 
 
-def _handle_successful_response(browser: Browser, response: Response, url: str,
-                                history: bool):
+def _handle_successful_response(browser: Browser, response: Response, url: str):
     """Handle a successful response content from a Gemini server.
 
     According to the MIME type received or inferred, the response is either
@@ -150,8 +153,11 @@ def _handle_successful_response(browser: Browser, response: Response, url: str,
     - browser: Browser instance that made the initial request.
     - url: original URL.
     - response: a successful Response.
-    - history: whether to modify history on a page load.
+
+    Returns:
+    True on success, False otherwise.
     """
+    # Use appropriate response parser according to the MIME type.
     mime_type = response.get_mime_type()
     page = None
     error = None
@@ -171,13 +177,14 @@ def _handle_successful_response(browser: Browser, response: Response, url: str,
     else:
         filepath = _get_download_path(url)
 
+    # If a page has been produced, load it. Else if a file has been retrieved,
+    # download it.
     if page:
         browser.load_page(page)
-        if browser.current_url and history:
-            browser.history.push(browser.current_url)
         browser.current_url = url
         browser.cache[url] = page
         browser.set_status(url)
+        return True
     elif filepath:
         try:
             with open(filepath, "wb") as download_file:
@@ -186,8 +193,11 @@ def _handle_successful_response(browser: Browser, response: Response, url: str,
             browser.set_status_error(f"Failed to save {url} ({exc})")
         else:
             browser.set_status(f"Downloaded {url} ({mime_type.short}).")
+            browser.last_download = mime_type, filepath
+            return True
     elif error:
         browser.set_status_error(error)
+    return False
 
 
 def _get_download_path(url: str) -> Path:

bebop/config.py

@@ -9,6 +9,9 @@ DEFAULT_CONFIG = {
     "text_width": 80,
     "source_editor": ["vi"],
     "command_editor": ["vi"],
+    "history_limit": 1000,
+    "external_commands": {},
+    "external_command_default": ["xdg-open"]
 }
 
 

bebop/help.py

@@ -1,38 +1,49 @@
 """Help page. Currently only keybinds are shown as help."""
 
 HELP_PAGE = """\
-# Bebop keybinds
+# Help
+
+## Keybinds
 
 Keybinds using the SHIFT key are written uppercase. Keybinds using the ALT (or \
 META) key are written using the "M-" prefix. Symbol keys are written as their \
 name, not the symbol itself.
 
-``` list of bebop keybinds
-- colon: focus the command-line
-- r: reload page
-- h: scroll left a bit
-- j: scroll down a bit
-- k: scroll up a bit
-- l: scroll right a bit
-- H: scroll left a whole page
-- J: scroll down a whole page
-- K: scroll up a whole page
-- L: scroll right a whole page
-- M-h: scroll one column left
-- M-j: scroll one line down
-- M-k: scroll one line up
-- M-l: scroll one column right
-- circumflex: horizontally scroll back to the first column
-- gg: go to the top of the page
-- G: go to the bottom of the page
-- o: open an URL
-- p: go to the previous page
-- u: go to the parent page (up a level in URL)
-- U: go to the root page (root URL for the current domain)
-- b: open bookmarks
-- B: add current page to bookmarks
-- e: open the current page source in an editor
-- digits: go to the corresponding link ID
-- escape: reset status line text
-```
+* colon: focus the command-line
+* r: reload page
+* h: scroll left a bit
+* j: scroll down a bit
+* k: scroll up a bit
+* l: scroll right a bit
+* H: scroll left a whole page
+* J: scroll down a whole page
+* K: scroll up a whole page
+* L: scroll right a whole page
+* M-h: scroll one column left
+* M-j: scroll one line down
+* M-k: scroll one line up
+* M-l: scroll one column right
+* circumflex: horizontally scroll back to the first column
+* gg: go to the top of the page
+* G: go to the bottom of the page
+* o: open an URL
+* O: open last download with an external command
+* p: go to the previous page
+* u: go to the parent page (up a level in URL)
+* U: go to the root page (root URL for the current domain)
+* b: open bookmarks
+* B: add current page to bookmarks
+* e: open the current page source in an editor
+* y: open history
+* digits: go to the corresponding link ID
+* escape: reset status line text
+
+## Commands
+
+Commands are mostly for actions requiring user input. You can type a command \
+with arguments by pressing the corresponding keybind above.
+
+* o/open <url>: open this URL
+* forget_certificate <hostname>: remove saved fingerprint for the hostname
+* q/quit: well, quit
 """

bebop/history.py

@@ -2,20 +2,56 @@
 
 
 class History:
-    """Basic browsing history manager."""
+    """Basic browsing history manager.
 
-    def __init__(self):
+    """
+
+    def __init__(self, limit):
         self.urls = []
-
-    def has_links(self):
-        """Return True if there is at least one URL in the history."""
-        return bool(self.urls)
+        self.backlist = []
+        self.limit = limit
 
     def push(self, url):
         """Add an URL to the history."""
-        if not self.urls or self.urls[-1] != url:
+        # Append url to our URLs, bubbling it up if it's already there.
+        try:
+            self.urls.remove(url)
+        except ValueError:
+            pass
         self.urls.append(url)
+        if len(self.urls) > self.limit:
+            self.urls.pop(0)
 
-    def pop(self):
-        """Return latest URL added to history and remove it."""
-        return self.urls.pop()
+        # Also simply push it to the backlist.
+        self.backlist.append(url)
+        if len(self.backlist) > self.limit:
+            self.backlist.pop(0)
+
+    def get_previous(self, actual_previous=False):
+        """Return previous URL, or None if there is only one or zero URL."""
+        try:
+            if actual_previous:
+                return self.backlist[-1]
+            # The backlist should be populated with the first link visited and
+            # never completely emptied afterwards, or we end up in situation
+            # where you can't get away from internal pages.
+            if len(self.backlist) > 1:
+                self.backlist.pop()
+            return self.backlist[-1]
+        except IndexError:
+            return None
+
+    def to_gemtext(self):
+        """Generate a simple Gemtext page of the current history.
+
+        Present a page that follows the "by last visited" behaviour of Firefox
+        for the lack of a better idea, avoiding duplicated entries.
+        """
+        urls = []
+        seen = set()
+        for url in reversed(self.urls):
+            if url in seen:
+                continue
+            urls.append(url)
+            seen.add(url)
+        return "# History\n\n" + "\n".join("=> " + url for url in urls)

bebop/navigation.py

@@ -1,66 +1,191 @@
-"""URI (RFC 3986) helpers for Gemini navigation."""
+"""URI (RFC 3986) helpers for Gemini navigation.
 
-import urllib.parse
+It was supposed to be just thin fixes around urllib.parse functions but as
+gemini is not recognized as a valid scheme it breaks a lot of things, so it
+turned into a basic re-implementation of the RFC.
+"""
+
+import re
+from typing import Any, Dict, Optional
+from urllib.parse import quote
 
 
-def parse_url(url: str, absolute: bool =False):
+URI_RE = re.compile(
+    "^"
+    r"(?:(?P<scheme>[^:/?#\n]+):)?"
+    r"(?://(?P<netloc>[^/?#\n]*))?"
+    r"(?P<path>[^?#\n]*)"
+    r"(?:\?(?P<query>[^#\n]*))?"
+    r"(?:#(?P<fragment>.*))?"
+    "$"
+)
+
+
+class InvalidUrlException(Exception):
+    """Generic exception for invalid URLs used in this module."""
+
+    def __init__(self, url):
+        super().__init__()
+        self.url = url
+
+
+def parse_url(
+    url: str,
+    absolute: bool =False,
+    default_scheme: Optional[str] =None
+) -> Dict[str, Any]:
     """Return URL parts from this URL.
 
-    This uses urllib.parse.urlparse to not reinvent the wheel, with a few
-    adjustments.
+    Use the RFC regex to get parts from URL. This function can be used on
+    regular URLs but also on not-so-compliant URLs, e.g. "capsule.org/page",
+    which might be typed by an user (see `absolute` argument).
 
-    First, urllib does not know the Gemini scheme (yet!) so if it
-    is specified we strip it to get an absolute netloc.
+    Arguments:
+    - url: URL to parse.
+    - absolute: assume the URL is absolute, e.g. in the case we are trying to
+      parse an URL an user has written, which is most of the time an absolute
+      URL even if not perfectly so. This only has an effect if, after the
+      initial parsing, there is no scheme or netloc available.
+    - default_scheme: specify the scheme to use if the URL either does not
+      specify it and we need it (e.g. there is a location), or `absolute` is
+      true; if absolute is true but `default_scheme` is not specified, use the
+      gemini scheme.
 
-    Second, as this function can be used to process arbitrary user input, we
-    clean it a bit:
-    - strip whitespaces from the URL
-    - if "absolute" is True, consider that the URL is meant to be absolute, even
-      though it technically is not, e.g. "dece.space" is not absolute as it
-      misses either the // delimiter.
+    Returns:
+    URL parts, as a dictionary with the following keys: "scheme", "netloc",
+    "path", "query" and "fragment". All keys are present, but all values can be
+    None, except path which is always a string (but can be empty).
+
+    Raises:
+    InvalidUrlException if you put really really stupid strings in there.
     """
-    url = url.strip()
-    if url.startswith("file://"):
-        return urllib.parse.urlparse(url)
-    if url.startswith("gemini://"):
-        url = url[7:]
-    parts = urllib.parse.urlparse(url, scheme="gemini")
-    if not parts.netloc or absolute:
-        parts = urllib.parse.urlparse(f"//{url}", scheme="gemini")
+    match = URI_RE.match(url)
+    if not match:
+        raise InvalidUrlException(url)
+
+    match_dict = match.groupdict()
+    parts = {
+        k: match_dict.get(k)
+        for k in ("scheme", "netloc", "path", "query", "fragment")
+    }
+
+    # Smol hack: if we assume it's an absolute URL, just prefix scheme and "//".
+    if absolute and not parts["scheme"] and not parts["netloc"]:
+        scheme = default_scheme or "gemini"
+        return parse_url(scheme + "://" + url)
+
+    # Another smol hack: if there is no scheme, use `default_scheme` as default.
+    if default_scheme and parts["scheme"] is None:
+        parts["scheme"] = default_scheme
+
     return parts
 
 
-def sanitize_url(url: str):
-    """Parse and unparse an URL to ensure it has been properly formatted."""
-    return urllib.parse.urlunparse(parse_url(url))
+def unparse_url(parts) -> str:
+    """Unparse parts of an URL produced by `parse_url`."""
+    url = ""
+    if parts["scheme"] is not None:
+        url += parts["scheme"] + ":"
+    if parts["netloc"] is not None:
+        url += "//" + parts["netloc"]
+    if parts["path"] is not None:
+        url += parts["path"]
+    if parts["query"] is not None:
+        url += "?" + parts["query"]
+    if parts["fragment"] is not None:
+        url += "#" + parts["fragment"]
+    return url
 
 
-def join_url(base_url: str, url: str):
-    """Join a base URL with a relative url."""
-    if base_url.startswith("gemini://"):
-        base_url = base_url[7:]
-    parts = parse_url(urllib.parse.urljoin(base_url, url))
-    return urllib.parse.urlunparse(parts)
+def clear_post_path(parts) -> None:
+    """Clear optional post-path parts (query and fragment)."""
+    parts["query"] = None
+    parts["fragment"] = None
 
 
-def set_parameter(url: str, user_input: str):
+def join_url(base_url: str, rel_url: str) -> str:
+    """Join a base URL with a relative path."""
+    parts = parse_url(base_url)
+    rel_parts = parse_url(rel_url)
+    if rel_url.startswith("/"):
+        new_path = rel_parts["path"]
+    else:
+        base_path = parts["path"] or ""
+        new_path = remove_last_segment(base_path) + "/" + rel_parts["path"]
+    parts["path"] = remove_dot_segments(new_path)
+    parts["query"] = rel_parts["query"]
+    parts["fragment"] = rel_parts["fragment"]
+    return unparse_url(parts)
+
+
+def remove_dot_segments(path: str):
+    """Remove dot segments in an URL path."""
+    output = ""
+    while path:
+        if path.startswith("../"):
+            path = path[3:]
+        elif path.startswith("./") or path.startswith("/./"):
+            path = path[2:]  # Either strip "./" or leave a single "/".
+        elif path == "/.":
+            path = "/"
+        elif path.startswith("/../"):
+            path = "/" + path[4:]
+            output = remove_last_segment(output)
+        elif path == "/..":
+            path = "/"
+            output = remove_last_segment(output)
+        elif path in (".", ".."):
+            path = ""
+        else:
+            first_segment, path = pop_first_segment(path)
+            output += first_segment
+    return output
+
+
+def remove_last_segment(path: str):
+    """Remove last path segment, including preceding "/" if any."""
+    return path[:path.rfind("/")]
+
+
+def pop_first_segment(path: str):
+    """Return first segment and the rest.
+
+    Return the first segment including the initial "/" if any, and the rest of
+    the path up to, but not including, the next "/" or the end of the string.
+    """
+    next_slash = path[1:].find("/")
+    if next_slash == -1:
+        return path, ""
+    next_slash += 1
+    return path[:next_slash], path[next_slash:]
+
+
+def set_parameter(url: str, user_input: str) -> str:
     """Return a new URL with the escaped user input appended."""
-    quoted_input = urllib.parse.quote(user_input)
-    if "?" in url:
-        url = url.split("?", maxsplit=1)[0]
-    return url + "?" + quoted_input
+    parts = parse_url(url)
+    parts["query"] = quote(user_input)
+    return unparse_url(parts)
+
+
+def get_parent_path(path: str) -> str:
+    """Return the parent path."""
+    last_slash = path.rstrip("/").rfind("/")
+    if last_slash > -1:
+        path = path[:last_slash + 1]
+    return path
 
 
 def get_parent_url(url: str) -> str:
     """Return the parent URL (one level up)."""
-    scheme, netloc, path, _, _, _ = parse_url(url)
-    last_slash = path.rstrip("/").rfind("/")
-    if last_slash > -1:
-        path = path[:last_slash + 1]
-    return urllib.parse.urlunparse((scheme, netloc, path, "", "", ""))
+    parts = parse_url(url)
+    parts["path"] = get_parent_path(parts["path"])  # type: ignore
+    clear_post_path(parts)
+    return unparse_url(parts)
 
 
 def get_root_url(url: str) -> str:
     """Return the root URL (basically discards path)."""
-    scheme, netloc, _, _, _, _ = parse_url(url)
-    return urllib.parse.urlunparse((scheme, netloc, "/", "", "", ""))
+    parts = parse_url(url)
+    parts["path"] = "/"
+    clear_post_path(parts)
+    return unparse_url(parts)

bebop/rendering.py

@@ -15,7 +15,7 @@ def render_lines(metalines, window, max_width):
 
     Arguments:
     - metalines: list of metalines to render, must have at least one element.
-    - window: window that will be resized as filled with rendered lines.
+    - window: window that will be resized and filled with rendered lines.
     - max_width: line length limit for the pad.
 
     Returns:
@@ -24,7 +24,7 @@ def render_lines(metalines, window, max_width):
     dimensions of the resized window.
     """
     num_lines = len(metalines)
-    new_dimensions = num_lines, max_width
+    new_dimensions = max(num_lines, 1), max_width
     window.resize(*new_dimensions)
     for line_index, metaline in enumerate(metalines):
         try:

bebop/tests/test_navigation.py

@@ -1,32 +1,54 @@
 import unittest
 
-from ..navigation import join_url, parse_url, set_parameter
+from ..navigation import (
+    get_parent_url, get_root_url, join_url, parse_url, pop_first_segment, remove_dot_segments,
+    remove_last_segment, set_parameter,
+)
 
 
 class TestNavigation(unittest.TestCase):
 
     def test_parse_url(self):
-        res = parse_url("gemini://dece.space/parse-me.gmi")
-        self.assertEqual(res.scheme, "gemini")
-        self.assertEqual(res.netloc, "dece.space")
-        self.assertEqual(res.path, "/parse-me.gmi")
+        # Basic complete URL.
+        res = parse_url("gemini://netloc/parse-me.gmi")
+        self.assertEqual(res["scheme"], "gemini")
+        self.assertEqual(res["netloc"], "netloc")
+        self.assertEqual(res["path"], "/parse-me.gmi")
 
-        res_netloc = parse_url("//dece.space/parse-me.gmi")
-        self.assertEqual(res, res_netloc)
+        # No scheme.
+        res_netloc = parse_url("//netloc/parse-me.gmi")
+        self.assertIsNone(res_netloc["scheme"], None)
+        for key in res_netloc:
+            if key == "scheme":
+                continue
+            self.assertEqual(res_netloc[key], res[key])
 
+        # No scheme but a default is provided.
+        res_netloc = parse_url("//netloc/parse-me.gmi", default_scheme="gemini")
+        self.assertDictEqual(res_netloc, res)
+
+        # No scheme nor netloc: only a path should be produced.
+        res = parse_url("dece.space/parse-me.gmi")
+        self.assertIsNone(res["scheme"])
+        self.assertIsNone(res["netloc"])
+        self.assertEqual(res["path"], "dece.space/parse-me.gmi")
+
+        # No scheme nor netloc but we should pretend having an absolute URL.
         res = parse_url("dece.space/parse-me.gmi", absolute=True)
-        self.assertEqual(res.scheme, "gemini")
-        self.assertEqual(res.netloc, "dece.space")
-        self.assertEqual(res.path, "/parse-me.gmi")
+        self.assertEqual(res["scheme"], "gemini")
+        self.assertEqual(res["netloc"], "dece.space")
+        self.assertEqual(res["path"], "/parse-me.gmi")
 
+        # HTTPS scheme.
         res = parse_url("https://dece.space/index.html")
-        self.assertEqual(res.scheme, "https")
-        self.assertEqual(res.netloc, "dece.space")
-        self.assertEqual(res.path, "/index.html")
+        self.assertEqual(res["scheme"], "https")
+        self.assertEqual(res["netloc"], "dece.space")
+        self.assertEqual(res["path"], "/index.html")
 
+        # File scheme.
         res = parse_url("file:///home/dece/gemini/index.gmi")
-        self.assertEqual(res.scheme, "file")
-        self.assertEqual(res.path, "/home/dece/gemini/index.gmi")
+        self.assertEqual(res["scheme"], "file")
+        self.assertEqual(res["path"], "/home/dece/gemini/index.gmi")
 
     def test_join_url(self):
         url = join_url("gemini://dece.space/", "some-file.gmi")
@@ -39,9 +61,84 @@ class TestNavigation(unittest.TestCase):
         self.assertEqual(url, "gemini://dece.space/dir1/other-file.gmi")
         url = join_url("gemini://dece.space/dir1/file.gmi", "../top-level.gmi")
         self.assertEqual(url, "gemini://dece.space/top-level.gmi")
+        url = join_url("s://hard/dir/a", "./../test/b/c/../d/e/f/../.././a.gmi")
+        self.assertEqual(url, "s://hard/test/b/d/a.gmi")
+
+    def test_remove_dot_segments(self):
+        paths = [
+            ("index.gmi", "index.gmi"),
+            ("/index.gmi", "/index.gmi"),
+            ("./index.gmi", "index.gmi"),
+            ("/./index.gmi", "/index.gmi"),
+            ("/../index.gmi", "/index.gmi"),
+            ("/a/b/c/./../../g", "/a/g"),
+            ("mid/content=5/../6", "mid/6"),
+            ("../../../../g", "g"),
+        ]
+        for path, expected in paths:
+            self.assertEqual(
+                remove_dot_segments(path),
+                expected,
+                msg="path was " + path
+            )
+
+    def test_remove_last_segment(self):
+        self.assertEqual(remove_last_segment(""), "")
+        self.assertEqual(remove_last_segment("/"), "")
+        self.assertEqual(remove_last_segment("/a"), "")
+        self.assertEqual(remove_last_segment("/a/"), "/a")
+        self.assertEqual(remove_last_segment("/a/b"), "/a")
+        self.assertEqual(remove_last_segment("/a/b/c/d"), "/a/b/c")
+        self.assertEqual(remove_last_segment("///"), "//")
+
+    def test_pop_first_segment(self):
+        self.assertEqual(pop_first_segment(""), ("", ""))
+        self.assertEqual(pop_first_segment("a"), ("a", ""))
+        self.assertEqual(pop_first_segment("/a"), ("/a", ""))
+        self.assertEqual(pop_first_segment("/a/"), ("/a", "/"))
+        self.assertEqual(pop_first_segment("/a/b"), ("/a", "/b"))
+        self.assertEqual(pop_first_segment("a/b"), ("a", "/b"))
 
     def test_set_parameter(self):
         url = set_parameter("gemini://gus.guru/search", "my search")
         self.assertEqual(url, "gemini://gus.guru/search?my%20search")
         url = set_parameter("gemini://gus.guru/search?old%20search", "new")
         self.assertEqual(url, "gemini://gus.guru/search?new")
+
+    def test_get_parent_url(self):
+        urls_and_parents = [
+            ("gemini://host", "gemini://host"),
+            ("gemini://host/", "gemini://host/"),
+            ("gemini://host/a", "gemini://host/"),
+            ("gemini://host/a/", "gemini://host/"),
+            ("gemini://host/a/index.gmi", "gemini://host/a/"),
+            ("gemini://host/a/b/", "gemini://host/a/"),
+            ("gemini://host/a/b/file.flac", "gemini://host/a/b/"),
+            ("//host/a/b", "//host/a/"),
+            ("hey", "hey"),  # does not really make sense but whatever
+            ("hey/ho", "hey/"),
+            ("hey/ho/letsgo", "hey/ho/"),
+        ]
+        for url, parent in urls_and_parents:
+            self.assertEqual(
+                get_parent_url(url),
+                parent,
+                msg="URL was " + url)
+
+    def test_get_root_url(self):
+        urls_and_roots = [
+            ("gemini://host", "gemini://host/"),
+            ("gemini://host/", "gemini://host/"),
+            ("gemini://host/a", "gemini://host/"),
+            ("gemini://host/a/b/c", "gemini://host/"),
+            ("//host/path", "//host/"),
+            ("//host/path?query", "//host/"),
+            ("dumb", "/"),
+            ("dumb/dumber", "/"),
+        ]
+        for url, root in urls_and_roots:
+            self.assertEqual(
+                get_root_url(url), 
+                root, 
+                msg="URL was " + url
+            )