Implement merging Cargo.toml file with workspace file

easybuild/easyblocks/generic/cargo.py

@@ -37,6 +37,7 @@
 import tempfile
 from glob import glob
 from pathlib import Path
+from typing import Dict, List, Union
 
 import easybuild.tools.environment as env
 import easybuild.tools.systemtools as systemtools
@@ -78,19 +79,136 @@
 CARGO_CHECKSUM_JSON = '{{"files": {{}}, "package": "{checksum}"}}'
 
 
-def get_workspace_members(crate_dir: Path):
-    """Find all members of a cargo workspace in crate_dir.
+def parse_toml_list(value: str) -> List[str]:
+    """Split a TOML list value"""
+    if not value.startswith('[') or not value.endswith(']'):
+        raise ValueError(f"'{value}' is not a TOML list")
+    value = value[1:-1].strip()
+    simple_str_markers = ('"""', "'''", "'")
+    current_value = ''
+    result = []
+    while value:
+        for marker in simple_str_markers:
+            if value.startswith(marker):
+                idx = value.index(marker, len(marker))
+                current_value += value[:idx + len(marker)]
+                value = value[idx + len(marker):].lstrip()
+                break
+        else:
+            if value.startswith('"'):
+                m = re.match(r'".*?(?<!\\)"', value, re.M)
+                current_value += m[0]
+                value = value[m.end():].lstrip()
+        # Not inside a string here
+        if value.startswith(','):
+            result.append(current_value)
+            current_value = ''
+            value = value[1:].lstrip()
+        else:
+            m = re.search('"|\'|,', value)
+            if m:
+                current_value += value[:m.start()].strip()
+                value = value[m.end():]
+            else:
+                current_value += value.strip()
+                break
+    if current_value:
+        result.append(current_value)
+    return result
+
+
+def _clean_line(line: str, expected_end: Union[str, None]) -> str:
+    """Remove comments and trim line"""
+    if '#' not in line:
+        return line.strip()
+    if expected_end is not None and expected_end[0] in ("'", '"'):
+        try:
+            idx = line.index(expected_end) + len(expected_end)
+        except ValueError:
+            return line.strip()  # Ignore #-sign in multi-line string
+    else:
+        idx = 0
+    in_str = False
+    escaped = False
+    while idx < len(line):
+        c = line[idx]
+        if in_str:
+            if escaped:
+                if c == '\\':
+                    escaped = False
+            elif c == '"':
+                in_str = False
+            elif c == '\\':
+                escaped = True
+        elif c == '#':
+            break
+        elif c == '"':
+            in_str = True
+        elif c == "'":
+            try:
+                idx = line.index("'", idx + 1)
+            except ValueError:
+                idx = len(line)
+        idx += 1
+    return line[:idx].strip()
+
+
+def parse_toml(file_or_content: Union[Path, str]) -> Dict[str, str]:
+    """Minimally parse a TOML file into sections, keys and values
+
+    Values will be the raw strings (including quotes for string-typed values)"""
+
+    result: Dict[str, Union[str, List[str]]] = {}
+    pending_key = None
+    pending_value = None
+    expected_end = None
+    current_section = None
+    content = read_file(file_or_content) if isinstance(file_or_content, Path) else file_or_content
+    line_num = raw_line = None
+    start_end = {
+        '[': ']',
+        '{': '}',
+        '"""': '"""',
+        "'''": "'''",
+    }
+    try:
+        for line_num, raw_line in enumerate(content.splitlines()): # noqa B007: line_num used in error only
+            line: str = _clean_line(raw_line, expected_end)
+            if not line:
+                continue
+            if pending_key is None and line.startswith("[") and line.endswith("]"):
+                current_section = line.strip()[1:-1].strip()
+                result.setdefault(current_section, {})
+                continue
+            if pending_key is None:
+                key, val = line.split("=", 1)
+                pending_key = key.strip()
+                pending_value = val.strip()
+                for start, end in start_end.items():
+                    if pending_value.startswith(start):
+                        expected_end = end
+                        break
+                else:
+                    expected_end = None
+            else:
+                pending_value += '\n' + line
+            if expected_end is None or (pending_value != expected_end and pending_value.endswith(expected_end)):
+                result[current_section][pending_key] = pending_value.strip()
+                pending_key = None
+    except Exception as e:
+        raise ValueError(f'Failed to parse {file_or_content}, error {e} at line {line_num}: {raw_line}')
+    return result
 
-    (Minimally) parse the Cargo.toml file.
+
+def get_workspace_members(cargo_toml: Dict[str, str]):
+    """Find all members of a cargo workspace in the parsed the Cargo.toml file.
 
     Return a tuple: (has_package, workspace-members).
     has_package determines if it is a virtual workspace ([workspace] and no [package])
     workspace-members are all members (subfolder names) if it is a workspace, otherwise None
     """
-    cargo_toml = crate_dir / 'Cargo.toml'
-    lines = [line.strip() for line in read_file(cargo_toml).splitlines()]
     # A virtual (workspace) manifest has no [package], but only a [workspace] section.
-    has_package = '[package]' in lines
+    has_package = 'package' in cargo_toml
 
     # We are looking for this:
     # [workspace]
@@ -101,30 +219,15 @@ def get_workspace_members(crate_dir: Path):
     # ]
 
     try:
-        start_idx = lines.index('[workspace]')
-    except ValueError:
+        workspace = cargo_toml['workspace']
+    except KeyError:
         return has_package, None
-    # Find "members = [" and concatenate the value, stop at end of section or file
-    member_str = None
-    for line in lines[start_idx + 1:]:
-        if line.startswith('#'):
-            continue  # Skip comments
-        if re.match(r'\[\w+\]', line):
-            break  # New section
-        if member_str is None:
-            m = re.match(r'members\s+=\s+\[', line)
-            if m:
-                member_str = line[m.end():]
-        else:
-            member_str += line
-        # Stop if we reach the end of the list
-        if member_str is not None and member_str.endswith(']'):
-            member_str = member_str[:-1]
-            break
-    if member_str is None:
+    try:
+        member_strs = parse_toml_list(workspace['members'])
+    except (KeyError, ValueError):
         raise EasyBuildError('Failed to find members in %s', cargo_toml)
-    # Split at commas after removing possibly trailing ones and remove the quotes
-    members = [member.strip().strip('"') for member in member_str.rstrip(',').split(',')]
+    # Remove the quotes
+    members = [member.strip('"') for member in member_strs]
     # Sanity check that we didn't pick up anything unexpected
     invalid_members = [member for member in members if not re.match(r'(\w|-)+', member)]
     if invalid_members:
@@ -133,6 +236,35 @@ def get_workspace_members(crate_dir: Path):
     return has_package, members
 
 
+def merge_sub_crate(cargo_toml_path: Path, workspace_toml: Dict[str, str]):
+    """Resolve workspace references in the Cargo.toml file"""
+    # Lines such as 'authors.workspace = true' must be replaced by 'authors = <value from workspace.package>'
+    content: str = read_file(cargo_toml_path)
+    SUFFIX = '.workspace'
+    if SUFFIX not in content:
+        return
+    cargo_toml = parse_toml(content)
+    lines = content.splitlines()
+
+    def do_replacement(section, workspace_section):
+        if not section or not workspace_section:
+            return
+
+        for key, value in section.items():
+            if key.endswith(SUFFIX) and value == 'true':
+                real_key = key[:-len(SUFFIX)]
+                value = workspace_section[real_key]
+                idx = next(idx for idx, line in enumerate(lines) if key in line)
+                lines[idx] = f'{real_key} = {value}'
+
+    do_replacement(cargo_toml.get('package'), workspace_toml.get('workspace.package'))
+    do_replacement(cargo_toml.get('dependencies'), workspace_toml.get('workspace.dependencies'))
+    do_replacement(cargo_toml.get('build-dependencies'), workspace_toml.get('workspace.dependencies'))
+    do_replacement(cargo_toml.get('dev-dependencies'), workspace_toml.get('workspace.dependencies'))
+
+    write_file(cargo_toml_path, '\n'.join(lines))
+
+
 def get_checksum(src, log):
     """Get the checksum from an extracted source"""
     checksum = src['checksum']
@@ -354,7 +486,8 @@ def _setup_offline_config(self, git_sources):
         tmp_dir = Path(tempfile.mkdtemp(dir=self.builddir, prefix='tmp_crate_'))
         # Add checksum file for each crate such that it is recognized by cargo.
         # Glob to catch multiple folders in a source archive.
-        for crate_dir in (p.parent for p in Path(self.vendor_dir).glob('*/Cargo.toml')):
+        for cargo_toml in Path(self.vendor_dir).glob('*/Cargo.toml'):
+            crate_dir = cargo_toml.parent
             src = path_to_source.get(str(crate_dir))
             if src:
                 try:
@@ -372,7 +505,8 @@ def _setup_offline_config(self, git_sources):
             # otherwise (Only "[workspace]" section and no "[package]" section)
             # we have to remove the top-level folder or cargo fails with:
             # "found a virtual manifest at [...]Cargo.toml instead of a package manifest"
-            has_package, members = get_workspace_members(crate_dir)
+            parsed_toml = parse_toml(cargo_toml)
+            has_package, members = get_workspace_members(parsed_toml)
             if members:
                 self.log.info(f'Found workspace in {crate_dir}. Members: ' + ', '.join(members))
                 if not any((crate_dir / crate).is_dir() for crate in members):
@@ -397,6 +531,8 @@ def _setup_offline_config(self, git_sources):
                         # Use copy_dir to resolve symlinks that might point to the parent folder
                         copy_dir(tmp_crate_dir / member, target_path, symlinks=False)
                         cargo_pkg_dirs.append(target_path)
+                        self.log.info(f'Resolving workspace values for crate {member}')
+                        merge_sub_crate(target_path / 'Cargo.toml', parsed_toml)
                     if has_package:
                         # Remove the copied crate folders
                         for member in members: