Make jq errors cleaner for users when a header script is prepended. #…

…1414

pscheduler-core/pscheduler-core/validate-limits.raw

@@ -67,11 +67,9 @@ except IOError as ex:
 try:
     processor = LimitProcessor(infile)
 except Exception as ex:
-    pscheduler.fail(str(ex))
+    pscheduler.fail(f'Limit configuration is invalid:\n{str(ex)}')
 
 if sys.stdout.isatty() and not options.quiet:
     print("Limit configuration is valid.")
 
 pscheduler.succeed()
-
-

python-pscheduler/pscheduler/pscheduler/jqfilter.py

@@ -76,9 +76,9 @@ def __init__(
             args={},
             output_raw=False,
             groom=False,
+            strip_errors_to=None
             ):
-        """
-        Construct a filter.  Arguments:
+        """Construct a filter.  Arguments:
 
         filter_spec - The JQ script to be used for this filter.  This
         may be any subclass of basestring, a list or a dict.  Strings
@@ -95,6 +95,11 @@ def __init__(
         top of the script before compiling.  This allows filters
         prepending functions to user-provided scripts that do
         imports to compile properly.
+
+        strip_errors_to - Strip all lines in error messages up to the
+        string provided.  This is used to weed known-correct,
+        program-provided code out of errors on the first line so the
+        user sees the correct line numbers.
         """
 
         self.output_raw = output_raw
@@ -109,6 +114,9 @@ def __init__(
         if not isinstance(filter_spec, str):
             raise ValueError("Filter spec must be plain text, list or dict")
 
+        if strip_errors_to is not None and not isinstance(strip_errors_to, str):
+            raise ValueError("Strip string must be a string")
+
         # Passing an args hash into PyJQ causes a memory corruption
         # problem.  As a temporary workaround, turn them into "as"
         # statements.  (See #1059)
@@ -125,7 +133,28 @@ def __init__(
         if groom:
             filter_spec = _groom(filter_spec)
 
-        self.script = pyjq.compile(filter_spec, args, library_paths=_library_path())
+        value_error = None
+        try:
+            self.script = pyjq.compile(filter_spec, args, library_paths=_library_path())
+        except ValueError as ex:
+            # This is held and thrown seprately because Python will
+            # produce a confusing nested exception message when it's
+            # thrown.
+            value_error = ex
+
+        if value_error is not None:
+
+            if strip_errors_to is not None:
+                ex_lines = []
+                strip_length = len(strip_errors_to)
+                for num, line in enumerate(str(value_error).split('\n')):
+                    place = line.find(strip_errors_to)
+                    if place != -1:
+                        line = line[place+strip_length:]
+                    ex_lines.append(line.rstrip())
+                value_error = ValueError('\n'.join(ex_lines))
+
+            raise value_error
 
 
 

python-pscheduler/pscheduler/pscheduler/limitprocessor/rewriter.py

@@ -29,8 +29,7 @@ def __init__(self,
         else:
             script = transform["script"]
 
-
-        script_lines = [
+        SCRIPT_HEADER = [
 
             "def classifiers:",
             "  ." + self.PRIVATE_KEY + ".classifiers",
@@ -59,15 +58,22 @@ def __init__(self,
             "  error(\"Task rejected: \" + ($message | tostring))",
             ";",
 
-            script
-            ]
+            "def __END_HEADER__:",
+            "  .",
+            ";",
+            "__END_HEADER__ | "
+        ]
 
-        transform["script"] = script_lines
+        # Stuff the entire header onto the first line so errors
+        # reflect the line numbers of what the user provided.
+
+        transform["script"] = ' '.join(SCRIPT_HEADER) + script
 
         self.transform = JQFilter(
             filter_spec=transform,
             args=transform.get("args", {}),
-            groom=True
+            groom=True,
+            strip_errors_to='__END_HEADER__ | '
         )