to_full(): use namedtuple, improve sourcebuild version style

this picks some improvements from #275

setupsrc/pypdfium2_setup/build_pdfium.py

@@ -126,17 +126,17 @@ def identify_pdfium():
     # if not updated, we'll always be dirty because of the patches, so not much point checking it
     desc = run_cmd(["git", "describe", "--all"], cwd=PDFiumDir, capture=True)
     desc = desc.rsplit("/", maxsplit=1)[-1]
-    build, *id_parts = desc.split("-")
+    v_short, *id_parts = desc.split("-")
     assert len(id_parts) < 2
 
     # FIXME some duplication with base::parse_given_tag()
-    info = dict(build=build, n_commits=0, hash=None)
+    v_post = dict(n_commits=0, hash=None)
     if len(id_parts) > 0:
-        info["n_commits"] = int(id_parts[0])
+        v_post["n_commits"] = int(id_parts[0])
     if len(id_parts) > 1:
-        info["hash"] = id_parts[1]
+        v_post["hash"] = id_parts[1]
 
-    return info
+    return v_short, v_post
 
 
 def _create_resources_rc(pdfium_build):
@@ -170,14 +170,14 @@ def build(Ninja, target):
     run_cmd([Ninja, "-C", PDFiumBuildDir, target], cwd=PDFiumDir)
 
 
-def pack(pdfium_info):
+def pack(v_short, v_post):
 
     dest_dir = DataDir / ExtPlats.sourcebuild
     dest_dir.mkdir(parents=True, exist_ok=True)
 
     libname = LibnameForSystem[Host.system]
     shutil.copy(PDFiumBuildDir/libname, dest_dir/libname)
-    write_pdfium_info(dest_dir, origin="sourcebuild", **pdfium_info)
+    write_pdfium_info(dest_dir, v_short, origin="sourcebuild", **v_post)
 
     # We want to use local headers instead of downloading with build_pdfium_bindings(), therefore call run_ctypesgen() directly
     # FIXME PDFIUM_BINDINGS=reference not honored
@@ -237,10 +237,10 @@ def main(
     Ninja   = get_tool("ninja")
 
     pdfium_dl_done = dl_pdfium(GClient, b_update, b_revision)
-    pdfium_info = identify_pdfium()
+    v_short, v_post = identify_pdfium()
 
     if pdfium_dl_done:
-        patch_pdfium(pdfium_info["build"])
+        patch_pdfium(v_short)
     if b_use_syslibs:
         _dl_unbundler()
 
@@ -255,7 +255,7 @@ def main(
 
     configure(GN, config_str)
     build(Ninja, b_target)
-    pack(pdfium_info)
+    pack(v_short, v_post)
 
 
 def parse_args(argv):

setupsrc/pypdfium2_setup/craft_packages.py

@@ -168,7 +168,7 @@ def main_conda_bundle(args):
 
 def main_conda_raw(args):
     os.environ["PDFIUM_SHORT"] = str(args.pdfium_ver)
-    os.environ["PDFIUM_FULL"] = PdfiumVer.to_full(args.pdfium_ver, type=str)
+    os.environ["PDFIUM_FULL"] = ".".join([str(v) for v in PdfiumVer.to_full(args.pdfium_ver)])
     emplace_func = partial(prepare_setup, ExtPlats.system, args.pdfium_ver, use_v8=None)
     with CondaExtPlatfiles(emplace_func):
         run_conda_build(CondaDir/"raw", CondaDir/"raw"/"out")
@@ -179,7 +179,7 @@ def main_conda_helpers(args):
     # Set the current pdfium version as upper boundary, for inherent API safety.
     # Unfortunately, pdfium does not do semantic versioning, so it is hard to achieve safe upward flexibility.
     # See also https://groups.google.com/g/pdfium/c/kCmgW_gTFYE/m/BPoJgbwOCQAJ
-    # In case risk of conflicts becomes a problem, we could estimate an increase based on pdfium's deprecation period.
+    # In case the restrictive upper boundary becomes a problem, we could estimate an increase based on pdfium's deprecation period.
     # Relevant variables for such a calculation would be
     # - version increment speed (guess: average 2 per day)
     # - pdfium's lowest regular deprecation period (say: 6 months, as indicated by pdfium/CONTRIBUTING.md)

setupsrc/pypdfium2_setup/packaging_base.py

@@ -16,6 +16,7 @@
 import traceback
 import subprocess
 from pathlib import Path
+from collections import namedtuple
 import urllib.request as url_request
 
 # TODO(apibreak) consider renaming PDFIUM_PLATFORM to PDFIUM_BINARY ?
@@ -117,7 +118,7 @@ class ExtPlats:
 
 class PdfiumVer:
 
-    V_KEYS = ("major", "minor", "build", "patch")
+    scheme = namedtuple("PdfiumVer", ("major", "minor", "build", "patch"))
     _refs_cache = {"lines": None, "dict": {}, "cursor": None}
 
     @staticmethod
@@ -128,37 +129,33 @@ def get_latest():
         return int( tag.split("/")[-1] )
 
     @classmethod
-    def to_full(cls, v_short, type=dict):
+    def to_full(cls, v_short):
 
         # FIXME The ls-remote call is fairly expensive. While cached in memory for a process lifetime, it can cause a significant slowdown for consecutive process runs.
-        # There may be multiple ways to improve this, like adding a disk cache to ensure it would only be called once for a whole session, or adding a second strategy that would parse the pdfium-binaries VERSION file, and use the chromium refs only for sourcebuild.
+        # There may be multiple ways to improve this, like adding some disk cache to ensure it would only be called once for a whole session, or maybe adding a second strategy that would parse the pdfium-binaries VERSION file, and use the chromium refs only for sourcebuild.
 
         v_short = int(v_short)
         rc = cls._refs_cache
 
         if rc["lines"] is None:
+            print(f"Fetching chromium refs ...", file=sys.stderr)
             ChromiumURL = "https://chromium.googlesource.com/chromium/src"
             rc["lines"] = run_cmd(["git", "ls-remote", "--sort", "-version:refname", "--tags", ChromiumURL, '*.*.*.0'], cwd=None, capture=True).split("\n")
 
         if rc["cursor"] is None or rc["cursor"] > v_short:
             for i, line in enumerate(rc["lines"]):
                 ref = line.split("\t")[-1].rsplit("/", maxsplit=1)[-1]
-                major, minor, build, patch = [int(v) for v in ref.split(".")]
-                rc["dict"][build] = (major, minor, build, patch)
-                if build == v_short:
-                    rc["cursor"] = build
+                full_ver = cls.scheme(*[int(v) for v in ref.split(".")])
+                rc["dict"][full_ver.build] = full_ver
+                if full_ver.build == v_short:
+                    rc["cursor"] = full_ver.build
                     rc["lines"] = rc["lines"][i+1:]
                     break
 
-        v_parts = rc["dict"][v_short]
-        if type in (tuple, list):
-            return v_parts
-        elif type is str:
-            return ".".join([str(v) for v in v_parts])
-        elif type is dict:
-            return dict(zip(PdfiumVer.V_KEYS, v_parts))
-        else:
-            assert False
+        full_ver = rc["dict"][v_short]
+        print(f"Resolved {v_short} -> {full_ver}", file=sys.stderr)
+
+        return full_ver
 
 
 def read_json(fp):
@@ -171,7 +168,7 @@ def write_json(fp, data, indent=2):
 
 
 def write_pdfium_info(dir, build, origin, flags=[], n_commits=0, hash=None):
-    info = dict(**PdfiumVer.to_full(build, type=dict), n_commits=n_commits, hash=hash, origin=origin, flags=flags)
+    info = dict(**PdfiumVer.to_full(build)._asdict(), n_commits=n_commits, hash=hash, origin=origin, flags=flags)
     write_json(dir/VersionFN, info)
     return info