SaraView/Backend/Graph.py

"""The graph view's data (like Obsidian's graph): every directory and file
under one directory, joined to its parent directory, plus a link from a
file to every file or directory it references.

References are read from Markdown and text files:
- Markdown links and images: [Text](Other.md), ![](Images/Cat.png)
- Wiki links: [[Other]], [[Other|Shown]], ![[Cat.png]] (matched by name,
  like Obsidian, preferring the nearest one)
- Portals: {Path ***Name}, {Path head 3}, {Path '12'}

Listing follows the same rules as /api/tree (hidden files, .sv_ notes,
locks), so the graph never shows what the file list wouldn't. The desktop
client builds the same graph for local files (Client/Graph.py)."""

from __future__ import annotations

import re
from urllib.parse import unquote

from Core import NormalizeRelativePath, ResolvePath
from Explorer import ListDirectory
from Locks import IsPathAccessible
from Preview import ReadTextFile
from VarFunctions import MarkerPattern

# Past this, the graph is cut short (and says so) rather than taking ages
# to build and draw.
MaxNodes = 2000
# Bigger files aren't read for references.
MaxReadBytes = 1_000_000
ReferenceFileTypes = {"markdown", "text"}

FencedCodePattern = re.compile(r"^ {0,3}(```|~~~).*?^ {0,3}\1", re.MULTILINE | re.DOTALL)
MarkdownLinkPattern = re.compile(r"!?\[[^\]\n]*\]\(\s*<?([^)\s>]+)>?(?:\s+\"[^\"]*\")?\s*\)")
WikiLinkPattern = re.compile(r"!?\[\[([^\]|#\n]+)(?:#[^\]|\n]*)?(?:\|[^\]\n]*)?\]\]")
ExternalPattern = re.compile(r"^([a-z][a-z0-9+.-]*:|//|#)", re.IGNORECASE)


def FindReferences(Content: str) -> list[tuple[str, str]]:
    """("Path", relative path) and ("Name", wiki name) references in a file.
    Code blocks are skipped, since links in them are examples."""
    Content = FencedCodePattern.sub("", Content)
    Found: list[tuple[str, str]] = []
    for Match in MarkdownLinkPattern.finditer(Content):
        Target = Match.group(1)
        if ExternalPattern.match(Target):
            continue
        Target = unquote(Target.split("#", 1)[0].split("?", 1)[0])
        if Target:
            Found.append(("Path", Target))
    for Match in WikiLinkPattern.finditer(Content):
        Found.append(("Name", Match.group(1).strip()))
    for Match in MarkerPattern.finditer(Content):
        Target = Match.group("CrossPath") or Match.group("PipePath") or Match.group("LinePath")
        if Target:
            Found.append(("Path", Target))
    return Found


def ResolveReferences(
    References: list[tuple[str, str]], FileId: str, Ids: set[str], ByName: dict[str, list[str]]
) -> set[str]:
    """Node ids (paths relative to the root) that a file's references point to."""
    Directory = FileId.rsplit("/", 1)[0] if "/" in FileId else ""
    Targets: set[str] = set()
    for Kind, Target in References:
        if Kind == "Path":
            Base = "" if Target.startswith("/") else Directory
            Resolved = NormalizeRelativePath(Base, Target)
            if Resolved in Ids:
                Targets.add(Resolved)
            continue
        # Wiki link: a path ("Sub/Note"), else a name anywhere, nearest first.
        for Candidate in (Target, f"{Target}.md"):
            for Base in (Directory, ""):
                Resolved = NormalizeRelativePath(Base, Candidate)
                if Resolved in Ids:
                    Targets.add(Resolved)
                    break
            else:
                continue
            break
        else:
            Matches = ByName.get(Target.rsplit("/", 1)[-1].lower(), [])
            if Matches:
                Targets.add(min(Matches, key=lambda Id: (_Distance(Directory, Id), Id)))
    Targets.discard(FileId)
    return Targets


def _Distance(Directory: str, Id: str) -> int:
    """How many directories apart a file in Directory and Id are."""
    Here = Directory.split("/") if Directory else []
    There = Id.split("/")[:-1]
    Shared = 0
    for A, B in zip(Here, There):
        if A != B:
            break
        Shared += 1
    return len(Here) + len(There) - 2 * Shared


def IndexByName(Nodes: list[dict]) -> dict[str, list[str]]:
    """Files by lowercased name and by name without its extension, for wiki links."""
    ByName: dict[str, list[str]] = {}
    for Node in Nodes:
        if Node["IsDirectory"]:
            continue
        Name = Node["Name"].lower()
        ByName.setdefault(Name, []).append(Node["Id"])
        if "." in Name:
            ByName.setdefault(Name.rsplit(".", 1)[0], []).append(Node["Id"])
    return ByName


def BuildGraph(
    Root: str, Path: str, ShowHidden: bool, ShowSvFiles: bool, Username: str | None, UserGroups: list[str]
) -> dict:
    Path = NormalizeRelativePath("", Path)
    Nodes: list[dict] = [{"Id": Path, "Name": Path.rsplit("/", 1)[-1] if Path else Root,
                          "IsDirectory": True, "FileType": None}]
    Links: list[dict] = []
    Truncated = False
    Pending = [Path]

    while Pending and not Truncated:
        Directory = Pending.pop(0)
        try:
            Entries = ListDirectory(Root, Directory, ShowHidden, ShowSvFiles)
        except (ValueError, NotADirectoryError, FileNotFoundError, PermissionError):
            continue
        for Entry in Entries:
            Id = f"{Directory}/{Entry['Name']}" if Directory else Entry["Name"]
            if not IsPathAccessible(Root, Id, Username, UserGroups):
                continue
            if len(Nodes) >= MaxNodes:
                Truncated = True
                break
            Nodes.append({"Id": Id, "Name": Entry["Name"], "IsDirectory": Entry["IsDirectory"],
                          "FileType": Entry.get("FileType")})
            Links.append({"Source": Directory, "Target": Id, "Kind": "Child"})
            if Entry["IsDirectory"]:
                Pending.append(Id)

    Ids = {Node["Id"] for Node in Nodes}
    ByName = IndexByName(Nodes)
    for Node in Nodes:
        if Node["IsDirectory"] or Node["FileType"] not in ReferenceFileTypes:
            continue
        try:
            FilePath = ResolvePath(Root, Node["Id"])
            if FilePath.stat().st_size > MaxReadBytes:
                continue
            Content = ReadTextFile(FilePath)
        except (ValueError, OSError, UnicodeDecodeError):
            continue
        for Target in sorted(ResolveReferences(FindReferences(Content), Node["Id"], Ids, ByName)):
            Links.append({"Source": Node["Id"], "Target": Target, "Kind": "Reference"})

    return {"Root": Root, "Path": Path, "Nodes": Nodes, "Links": Links, "Truncated": Truncated}