0.5.0: DLM= preservation, JES2/3 cols 73-80 roundtrip, Python 3.13

DLM= custom delimiter:
- DD.instream_dlm field captures retain_delim from the scan tree
  (normalises /* and EMPTY_DELIM to None, keeps custom values)
- CSV serialiser adds instream_dlm column; rejcl reads it back
- rejcl emits DATA,DLM='XX' and the custom terminator on reverse path
- JSON/YAML include instream_dlm via asdict() when non-null
- Golden files updated for inlinedd.jcl which uses !! and zz delimiters

JES2/3 cols 73-80 roundtrip:
- _scan_jes2_control now calls _add_scanned_line so the tail (cols 72+)
  is captured; _reconstruct_jes2 appends tail[1:] (the 8-char sequence
  numbers) instead of padding with spaces
- _scan_jes3_control non-DATASET uses comment_text scan line matching
  the _scan_comment_stmt approach, fixing both roundtrip and emit paths

Quality:
- Python 3.13 classifier added; CI matrix extended to test both 3.12 and 3.13
- CLI smoke tests cover empty-bytes stdin warning and empty-file warning
- Regression tests cover instream_dlm model capture and JSON roundtrip
- CHANGELOG 0.5.0 entry covers all changes in this cycle

Author: msradam

.github/workflows/ci.yml

@@ -9,13 +9,16 @@ on:
 jobs:
   check:
     runs-on: ubuntu-latest
+    strategy:
+      matrix:
+        python-version: ["3.12", "3.13"]
     steps:
       - uses: actions/checkout@v4
 
       - name: Install uv
         uses: astral-sh/setup-uv@v5
         with:
-          python-version: "3.12"
+          python-version: ${{ matrix.python-version }}
 
       - name: Sync dependencies
         run: uv sync --all-groups

CHANGELOG.md

@@ -1,5 +1,26 @@
 # Changelog
 
+## 0.5.0 (unreleased)
+
+### Added
+
+- **`DD.instream_dlm`** field on the `DD` model: custom `DLM=` delimiters
+  are now preserved through the forward pass. JSON, YAML, and CSV
+  serialisers include the value; the rejcl reverse path emits `DATA,DLM='XX'`
+  and uses the custom terminator when reconstructing JCL.
+- Python 3.13 classifier and CI test matrix entry.
+
+### Fixed
+
+- **Scanner: JES2/3 cols 73-80 sequence numbers** on `/*` control statements
+  (e.g., `/*JOBPARM`, `/*SETUP`) were dropped during roundtrip. The scanner
+  now captures the tail, matching the `//` and `//*` behaviour.
+- **CLI: empty or comment-only JCL** now emits a warning to stderr with a
+  preview of the input (up to 5 lines) and exits 0 rather than silently
+  producing empty output.
+- **CLI: bare Python tracebacks** from I/O errors, bad paths, and serialiser
+  failures now surface as clean `fromjcl: <reason>` messages on stderr.
+
 ## 0.4.0 (2026-05-23)
 
 ### Added

pyproject.toml

@@ -18,6 +18,7 @@ classifiers = [
     "Operating System :: POSIX",
     "Programming Language :: Python :: 3",
     "Programming Language :: Python :: 3.12",
+    "Programming Language :: Python :: 3.13",
     "Topic :: Software Development :: Code Generators",
     "Topic :: System :: Systems Administration",
     "Topic :: Text Processing",
@@ -124,6 +125,7 @@ ignore_names = [
     "src_dataset",
     "region", "account", "programmer", "class_",
     "msgclass", "msglevel", "notify",
+    "instream_dlm",
 ]
 
 # --- Interrogate ---

src/fromjcl/_scanner.py

@@ -599,22 +599,28 @@ def _scan_comment_stmt(self, text: str) -> None:
         )
 
     def _scan_jes2_control(self, text: str) -> None:
-        # TODO: cols 73-80 sequence numbers are dropped; // and //* preserve them.
         self._add_stmt("/*", None)
         body_len = JCL_TXTLEN - PREFIX_LEN + 1
         key = text[PREFIX_LEN : PREFIX_LEN + body_len]
         self._add_kvp(key, None, None, True)
+        # Minimal scan line so _reconstruct_jes2 can recover cols 73-80 via tail.
+        self._add_scanned_line(None, None)
 
     def _scan_jes3_control(self, text: str, column: int) -> None:
-        # TODO: same cols 73-80 limitation as _scan_jes2_control above.
         self._add_stmt("//*", None)
         body = text[column:]
         if _word_starts(body, "DATASET"):
             self.state = ScanState.ContinueJES3Dataset
         else:
-            body_len = JCL_TXTLEN - PREFIX_LEN + 1
-            key = text[PREFIX_LEN : PREFIX_LEN + body_len]
-            self._add_kvp(key, None, None, True)
+            raw_len = len(self._current_raw)
+            content_end = max(column, min(raw_len, JCL_RECLEN))
+            comment_text = self._current_raw[column:content_end] or None
+            self._add_scanned_line(
+                None,
+                comment_text,
+                comment_col=column,
+                comment_end_col=column + len(comment_text or ""),
+            )
 
     def _dispatch(
         self,

src/fromjcl/models.py

@@ -193,6 +193,7 @@ class DD:
     sysout: str | None = None
     dummy: bool = False
     instream: str | None = None
+    instream_dlm: str | None = None
 
     @classmethod
     def from_statements(cls, stmts: list[dict[str, Any]]) -> "DD":
@@ -206,6 +207,9 @@ def from_statements(cls, stmts: list[dict[str, Any]]) -> "DD":
         inst = first.get("instream")
         instream = inst.get("bytes") if isinstance(inst, dict) else inst
         is_instream_dd = instream is not None
+        dlm = inst.get("retain_delim") if isinstance(inst, dict) else None
+        if not dlm or dlm in ("/*", "  "):
+            dlm = None
 
         for p in params:
             key = p["key"].upper() if p["key"] else ""
@@ -221,7 +225,7 @@ def from_statements(cls, stmts: list[dict[str, Any]]) -> "DD":
         if dummy:
             return cls(name=name, dummy=True)
         if is_instream_dd:
-            return cls(name=name, instream=instream or "")
+            return cls(name=name, instream=instream or "", instream_dlm=dlm)
 
         datasets = []
         for stmt in stmts:

src/fromjcl/rejcl.py

@@ -107,6 +107,8 @@ def _job_dict_from_csv(text: str) -> dict[str, Any]:
             elif row.get("instream"):
                 # Inverse of serialize/csv.py:_format_instream.
                 dd["instream"] = row["instream"].replace("\\n", "\n")
+                if row.get("instream_dlm"):
+                    dd["instream_dlm"] = row["instream_dlm"]
             else:
                 dd["datasets"] = []
             step["dds"].append(dd)
@@ -237,13 +239,17 @@ def _dd_statements(dd: dict[str, Any]) -> list[dict[str, Any]]:
             }
         ]
     if dd.get("instream") is not None:
-        # TODO: hardcoded `/*`. Custom DLM= is dropped on the forward pass.
+        dlm = dd.get("instream_dlm") or None
+        if dlm:
+            params = [{"key": "DATA", "value": None}, {"key": "DLM", "value": f"'{dlm}'"}]
+        else:
+            params = [{"key": "*", "value": None}]
         return [
             {
                 "type": "DD",
                 "name": name,
-                "parameters": [{"key": "*", "value": None}],
-                "instream": {"bytes": dd["instream"], "retain_delim": "/*"},
+                "parameters": params,
+                "instream": {"bytes": dd["instream"], "retain_delim": dlm or "/*"},
             }
         ]
     datasets = dd.get("datasets") or []

src/fromjcl/serialize/csv.py

@@ -42,6 +42,7 @@
     "sysout",
     "dummy",
     "instream",
+    "instream_dlm",
 ]
 
 
@@ -76,6 +77,7 @@ def _dd_base(step_base: dict[str, str], dd: DD) -> dict[str, str]:
         "sysout": dd.sysout or "",
         "dummy": "true" if dd.dummy else "",
         "instream": _format_instream(dd.instream),
+        "instream_dlm": dd.instream_dlm or "",
     }
 
 

src/fromjcl/serialize/jcl.py

@@ -371,6 +371,12 @@ def _reconstruct_jes2(stmt: dict[str, Any], target_len: int) -> str:
     # The scanner stores a 70-char body window; trim if the source record was shorter.
     if target_len and target_len < len(line):
         line = line[:target_len]
+    scan_lines = stmt.get("scanned_lines") or []
+    tail = (scan_lines[0].get("tail") or "") if scan_lines else ""
+    # body ends at col 72 (JCL_TXTLEN); tail starts at the same column,
+    # so tail[1:] (cols 73-80) are the sequence numbers to restore.
+    if len(tail) > 1:
+        return line + tail[1:]
     return _pad_to(line, target_len)
 
 

tests/jcl_samples/parser_edge_cases/SOURCES.md

@@ -22,5 +22,6 @@ The following files are hand-authored originals written for this project:
 
 | File                      | Notes |
 | ------------------------- | ----- |
-| `lowercase_jobname.jcl`   | Jobname containing lowercase characters (template placeholder pattern); exercises the name-char validation fix in 0.4.0 |
-| `acct_multi_element.jcl`  | JOB card with a parenthesised multi-element account field `(ACCT001,BIN1,BLDG2,DEPT3)`; exercises the paren-nesting fix in 0.4.0 |
+| `lowercase_jobname.jcl`      | Jobname containing lowercase characters (template placeholder pattern); exercises the name-char validation fix in 0.4.0 |
+| `acct_multi_element.jcl`     | JOB card with a parenthesised multi-element account field `(ACCT001,BIN1,BLDG2,DEPT3)`; exercises the paren-nesting fix in 0.4.0 |
+| `dlm_custom_delimiter.jcl`   | `DD DATA,DLM='@@'` with a custom instream terminator; exercises `instream_dlm` capture and JSON/YAML/CSV rejcl roundtrip |

tests/jcl_samples/parser_edge_cases/dlm_custom_delimiter.jcl

@@ -0,0 +1,7 @@
+//DLMTEST  JOB CLASS=A,MSGCLASS=A
+//STEP1    EXEC PGM=IEFBR14
+//SYSIN    DD DATA,DLM='@@'
+HELLO FROM INSTREAM
+ANOTHER RECORD
+@@
+//