def dot(self, fp, level_processes):
assert self.processes
if level_processes == LVL_PROC_RUN:
fp.write(' run%d [label="%d: %s"];\n' % (
self.nb, self.nb, self.processes[0].binary or "-"))
else:
fp.write(' subgraph cluster_run%d {\n label="%s";\n' % (
self.nb, escape(self.name)))
for process in self.processes:
if level_processes == LVL_PROC_THREAD or not process.thread:
process.dot(fp, level_processes, indent=2)
fp.write(' }\n')
def dot(self, fp, level_processes):
assert self.processes
if level_processes == LVL_PROC_RUN:
fp.write(' run%d [label="%d: %s"];\n' %
(self.nb, self.nb, self.processes[0].binary or "-"))
else:
fp.write(' subgraph cluster_run%d {\n label="%s";\n' %
(self.nb, escape(self.name)))
for process in self.processes:
if level_processes == LVL_PROC_THREAD or not process.thread:
process.dot(fp, level_processes, indent=2)
fp.write(' }\n')
def dot(self, fp, level_processes, indent=1):
thread_style = ',fillcolor="#666666"' if self.thread else ''
fp.write(' ' * indent + 'prog%d [label="%s (%d)"%s];\n' %
(self.id, escape(unicode_(self.binary) or "-"), self.pid,
thread_style))
if self.parent is not None:
reason = ''
if self.created == C_FORK:
if self.thread:
reason = "thread"
else:
reason = "fork"
elif self.created == C_EXEC:
reason = "exec"
elif self.created == C_FORKEXEC:
reason = "fork+exec"
fp.write(' ' * indent + 'prog%d -> prog%d [label="%s"];\n' %
(self.parent.id, self.id, reason))
def dot(self, fp, level_processes, indent=1):
thread_style = ',fillcolor="#666666"' if self.thread else ''
fp.write(' ' * indent + 'prog%d [label="%s (%d)"%s];\n' % (
self.id, escape(unicode_(self.binary) or "-"),
self.pid, thread_style))
if self.parent is not None:
reason = ''
if self.created == C_FORK:
if self.thread:
reason = "thread"
else:
reason = "fork"
elif self.created == C_EXEC:
reason = "exec"
elif self.created == C_FORKEXEC:
reason = "fork+exec"
fp.write(' ' * indent + 'prog%d -> prog%d [label="%s"];\n' % (
self.parent.id, self.id, reason))
def dot(self, fp, level_pkgs):
assert self.id is not None
if not self.files:
return
if level_pkgs == LVL_PKG_PACKAGE:
fp.write(' "pkg %s" [shape=box,label=' % escape(self.name))
if self.version:
fp.write('"%s %s"];\n' %
(escape(self.name), escape(self.version)))
else:
fp.write('"%s"];\n' % escape(self.name))
elif level_pkgs == LVL_PKG_FILE:
fp.write(' subgraph cluster_pkg%d {\n label=' % self.id)
if self.version:
fp.write('"%s %s";\n' %
(escape(self.name), escape(self.version)))
else:
fp.write('"%s";\n' % escape(self.name))
for f in sorted(unicode_(f) for f in self.files):
fp.write(' "%s";\n' % escape(f))
fp.write(' }\n')
def dot(self, fp, level_pkgs):
assert self.id is not None
if not self.files:
return
if level_pkgs == LVL_PKG_PACKAGE:
fp.write(' "pkg %s" [shape=box,label=' % escape(self.name))
if self.version:
fp.write('"%s %s"];\n' % (
escape(self.name), escape(self.version)))
else:
fp.write('"%s"];\n' % escape(self.name))
elif level_pkgs == LVL_PKG_FILE:
fp.write(' subgraph cluster_pkg%d {\n label=' % self.id)
if self.version:
fp.write('"%s %s";\n' % (
escape(self.name), escape(self.version)))
else:
fp.write('"%s";\n' % escape(self.name))
for f in sorted(unicode_(f) for f in self.files):
fp.write(' "%s";\n' % escape(f))
fp.write(' }\n')
def generate(target, configfile, database, all_forks=False):
"""Main function for the graph subcommand.
"""
# In here, a file is any file on the filesystem. A binary is a file, that
# gets executed. A process is a system-level task, identified by its pid
# (pids don't get reused in the database).
# What I call program is the couple (process, binary), so forking creates a
# new program (with the same binary) and exec'ing creates a new program as
# well (with the same process)
# Because of this, fork+exec will create an intermediate program that
# doesn't do anything (new process but still old binary). If that program
# doesn't do anything worth showing on the graph, it will be erased, unless
# all_forks is True (--all-forks).
# Reads package ownership from the configuration
if not configfile.is_file():
logging.critical("Configuration file does not exist!\n"
"Did you forget to run 'reprozip trace'?\n"
"If not, you might want to use --dir to specify an "
"alternate location.")
sys.exit(1)
runs, packages, other_files = load_config(configfile, canonical=False)
packages = dict((f.path, pkg) for pkg in packages for f in pkg.files)
if PY3:
# On PY3, connect() only accepts unicode
conn = sqlite3.connect(str(database))
else:
conn = sqlite3.connect(database.path)
conn.row_factory = sqlite3.Row
# This is a bit weird. We need to iterate on all types of events at the
# same time, ordering by timestamp, so we decorate-sort-undecorate
# Decoration adds timestamp (for sorting) and tags by event type, one of
# 'process', 'open' or 'exec'
# Reads processes from the database
process_cursor = conn.cursor()
process_rows = process_cursor.execute(
'''
SELECT id, parent, timestamp
FROM processes
ORDER BY id
''')
processes = {}
all_programs = []
# ... and opened files...
file_cursor = conn.cursor()
file_rows = file_cursor.execute(
'''
SELECT name, timestamp, mode, process
FROM opened_files
ORDER BY id
''')
binaries = set()
files = OrderedSet()
edges = OrderedSet()
# ... as well as executed files.
exec_cursor = conn.cursor()
exec_rows = exec_cursor.execute(
'''
SELECT name, timestamp, process, argv
FROM executed_files
ORDER BY id
''')
# Loop on all event lists
logging.info("Getting all events from database...")
rows = heapq.merge(((r[2], 'process', r) for r in process_rows),
((r[1], 'open', r) for r in file_rows),
((r[1], 'exec', r) for r in exec_rows))
for ts, event_type, data in rows:
if event_type == 'process':
r_id, r_parent, r_timestamp = data
if r_parent is not None:
parent = processes[r_parent]
binary = parent.binary
else:
parent = None
binary = None
p = Process(r_id,
parent,
r_timestamp,
False,
binary,
C_INITIAL if r_parent is None else C_FORK)
processes[r_id] = p
all_programs.append(p)
elif event_type == 'open':
r_name, r_timestamp, r_mode, r_process = data
r_name = PosixPath(r_name)
if r_mode != FILE_WDIR:
process = processes[r_process]
files.add(r_name)
edges.add((process, r_name, r_mode, None))
elif event_type == 'exec':
r_name, r_timestamp, r_process, r_argv = data
r_name = PosixPath(r_name)
process = processes[r_process]
binaries.add(r_name)
# Here we split this process in two "programs", unless the previous
# one hasn't done anything since it was created via fork()
if not all_forks and not process.acted:
process.binary = r_name
process.created = C_FORKEXEC
process.acted = True
else:
process = Process(process.pid,
process,
r_timestamp,
True, # Hides exec only once
r_name,
C_EXEC)
all_programs.append(process)
processes[r_process] = process
argv = tuple(r_argv.split('\0'))
if not argv[-1]:
argv = argv[:-1]
edges.add((process, r_name, None, argv))
process_cursor.close()
file_cursor.close()
conn.close()
# Puts files in packages
logging.info("Organizes packages...")
package_files = {}
other_files = []
for f in files:
pkg = packages.get(f)
if pkg is not None:
package_files.setdefault((pkg.name, pkg.version), []).append(f)
else:
other_files.append(f)
# Writes DOT file
with target.open('w', encoding='utf-8', newline='\n') as fp:
fp.write('digraph G {\n /* programs */\n node [shape=box];\n')
# Programs
logging.info("Writing programs...")
for program in all_programs:
fp.write(' prog%d [label="%s (%d)"];\n' % (
id(program), program.binary or "-", program.pid))
if program.parent is not None:
reason = ''
if program.created == C_FORK:
reason = "fork"
elif program.created == C_EXEC:
reason = "exec"
elif program.created == C_FORKEXEC:
reason = "fork+exec"
fp.write(' prog%d -> prog%d [label="%s"];\n' % (
id(program.parent), id(program), reason))
fp.write('\n node [shape=ellipse];\n\n /* system packages */\n')
# Files from packages
logging.info("Writing packages...")
for i, ((name, version), files) in enumerate(iteritems(package_files)):
fp.write(' subgraph cluster%d {\n label=' % i)
if version:
fp.write('"%s %s";\n' % (escape(name), escape(version)))
else:
fp.write('"%s";\n' % escape(name))
for f in files:
fp.write(' "%s";\n' % escape(unicode_(f)))
fp.write(' }\n')
fp.write('\n /* other files */\n')
# Other files
logging.info("Writing other files...")
for f in other_files:
fp.write(' "%s"\n' % escape(unicode_(f)))
fp.write('\n')
# Edges
logging.info("Connecting edges...")
for prog, f, mode, argv in edges:
if mode is None:
fp.write(' "%s" -> prog%d [color=blue, label="%s"];\n' % (
escape(unicode_(f)),
id(prog),
escape(' '.join(argv))))
elif mode & FILE_WRITE:
fp.write(' prog%d -> "%s" [color=red];\n' % (
id(prog), escape(unicode_(f))))
elif mode & FILE_READ:
fp.write(' "%s" -> prog%d [color=green];\n' % (
escape(unicode_(f)), id(prog)))
fp.write('}\n')
def graph_dot(target, runs, packages, other_files, package_map, edges,
inputs_outputs, level_pkgs, level_processes, level_other_files):
"""Writes a GraphViz DOT file from the collected information.
"""
with target.open('w', encoding='utf-8', newline='\n') as fp:
fp.write('digraph G {\n /* programs */\n'
' node [shape=box fontcolor=white '
'fillcolor=black style=filled];\n')
# Programs
logging.info("Writing programs...")
for run in runs:
run.dot(fp, level_processes)
fp.write('\n'
' node [shape=ellipse fontcolor="#131C39" '
'fillcolor="#C9D2ED"];\n')
# Packages
if level_pkgs not in (LVL_PKG_IGNORE, LVL_PKG_DROP):
logging.info("Writing packages...")
fp.write('\n /* system packages */\n')
for package in sorted(packages, key=lambda pkg: pkg.name):
package.dot(fp, level_pkgs)
fp.write('\n /* other files */\n')
# Other files
logging.info("Writing other files...")
for fi in sorted(other_files):
if fi in inputs_outputs:
fp.write(' "%(path)s" [fillcolor="#A3B4E0", '
'label="%(name)s\\n%(path)s"];\n' % {
'path': escape(unicode_(fi)),
'name': inputs_outputs[fi]
})
else:
fp.write(' "%s";\n' % escape(unicode_(fi)))
fp.write('\n')
# Edges
logging.info("Connecting edges...")
done_edges = set()
for prog, fi, mode, argv in edges:
endp_prog = prog.dot_endpoint(level_processes)
if fi in package_map:
if level_pkgs == LVL_PKG_DROP:
continue
endp_file = package_map[fi].dot_endpoint(fi, level_pkgs)
e = endp_prog, endp_file, mode
if e in done_edges:
continue
else:
done_edges.add(e)
else:
endp_file = '"%s"' % escape(unicode_(fi))
if mode is None:
fp.write(' %s -> %s [style=bold, label="%s"];\n' %
(endp_file, endp_prog, escape(format_argv(argv))))
elif mode & FILE_WRITE:
fp.write(' %s -> %s [color="#000088"];\n' %
(endp_prog, endp_file))
elif mode & FILE_READ:
fp.write(' %s -> %s [color="#8888CC"];\n' %
(endp_file, endp_prog))
fp.write('}\n')
def dot_endpoint(self, f, level_pkgs):
if level_pkgs == LVL_PKG_PACKAGE:
return '"pkg %s"' % escape(self.name)
else:
return '"%s"' % escape(unicode_(f))
def write_cltools_module(run, dot_vistrails):
input_files = run['input_files']
output_files = run['output_files']
module_name = 'reprounzip_%s' % hash_experiment_run(run)[:7]
# Writes CLTools JSON definition
(dot_vistrails / 'CLTools').mkdir(parents=True)
cltools_module = (dot_vistrails / 'CLTools' / module_name) + '.clt'
logging.info("Writing CLTools definition %s...", cltools_module)
with cltools_module.open('w', encoding='utf-8', newline='\n') as fp:
fp.write('{\n'
' "_comment": "This file was generated by reprounzip '
'%(version)s at %(date)s",\n\n' % {
'version': version,
'date': datetime.now().isoformat()
})
# python -m reprounzip.plugins.vistrails
fp.write(' "command": "%s",\n'
' "args": [\n'
' [\n'
' "constant",\n'
' "-m",\n'
' "flag",\n'
' {}\n'
' ],\n'
' [\n'
' "constant",\n'
' "reprounzip.plugins.vistrails",\n'
' "flag",\n'
' {}\n'
' ],\n' % escape(sys.executable))
# Unpacker
fp.write(' [\n'
' "input",\n'
' "unpacker",\n'
' "string",\n'
' {}\n'
' ],\n')
# Target directory
fp.write(' [\n'
' "input",\n'
' "directory",\n'
' "string",\n'
' {}\n'
' ],\n')
# Run number
fp.write(' [\n'
' "input",\n'
' "run",\n'
' "string",\n'
' {}\n'
' ],\n')
# Input files
for i, input_name in enumerate(input_files):
fp.write(' [\n'
' "input",\n'
' "input %(name)s",\n'
' "file",\n'
' {\n'
' "flag": "--input-file",\n'
' "prefix": "%(name)s:"\n'
' }\n'
' ],\n' % {'name': escape(input_name)})
# Output files
for i, output_name in enumerate(output_files):
fp.write(' [\n'
' "output",\n'
' "output %(name)s",\n'
' "file",\n'
' {\n'
' "flag": "--output-file",\n'
' "prefix": "%(name)s:"\n'
' }\n'
' ],\n' % {'name': escape(output_name)})
# Command-line
fp.write(' [\n'
' "input",\n'
' "cmdline",\n'
' "string",\n'
' {\n'
' "flag": "--cmdline"\n'
' }\n'
' ]\n'
' ],\n')
# Use "std file processing" since VisTrails <=2.1.4 has a bug without
# this (also, it's inefficient)
fp.write(' "options": {\n'
' "std_using_files": ""\n'
' },\n')
# Makes the module check for errors
fp.write(' "return_code": 0,\n')
# Enable 'stdout' port
fp.write(' "stdout": [\n'
' "stdout",\n'
' "file",\n'
' {}\n'
' ]\n'
'}\n')
return module_name
def generate(target, directory, all_forks=False):
"""Main function for the graph subcommand.
"""
# In here, a file is any file on the filesystem. A binary is a file, that
# gets executed. A process is a system-level task, identified by its pid
# (pids don't get reused in the database).
# What I call program is the couple (process, binary), so forking creates a
# new program (with the same binary) and exec'ing creates a new program as
# well (with the same process)
# Because of this, fork+exec will create an intermediate program that
# doesn't do anything (new process but still old binary). If that program
# doesn't do anything worth showing on the graph, it will be erased, unless
# all_forks is True (--all-forks).
database = directory / 'trace.sqlite3'
# Reads package ownership from the configuration
configfile = directory / 'config.yml'
if not configfile.is_file():
logging.critical("Configuration file does not exist!\n"
"Did you forget to run 'reprozip trace'?\n"
"If not, you might want to use --dir to specify an "
"alternate location.")
sys.exit(1)
runs, packages, other_files, patterns = load_config(configfile,
canonical=False)
packages = dict((f.path, pkg) for pkg in packages for f in pkg.files)
if PY3:
# On PY3, connect() only accepts unicode
conn = sqlite3.connect(str(database))
else:
conn = sqlite3.connect(database.path)
# This is a bit weird. We need to iterate on all types of events at the
# same time, ordering by timestamp, so we decorate-sort-undecorate
# Decoration adds timestamp (for sorting) and tags by event type, one of
# 'process', 'open' or 'exec'
# Reads processes from the database
process_cursor = conn.cursor()
process_rows = process_cursor.execute(
'''
SELECT id, parent, timestamp
FROM processes
ORDER BY id
''')
processes = {}
all_programs = []
# ... and opened files...
file_cursor = conn.cursor()
file_rows = file_cursor.execute(
'''
SELECT name, timestamp, mode, process
FROM opened_files
ORDER BY id
''')
binaries = set()
files = OrderedSet()
edges = OrderedSet()
# ... as well as executed files.
exec_cursor = conn.cursor()
exec_rows = exec_cursor.execute(
'''
SELECT name, timestamp, process, argv
FROM executed_files
ORDER BY id
''')
# Loop on all event lists
logging.info("Getting all events from database...")
rows = heapq.merge(((r[2], 'process', r) for r in process_rows),
((r[1], 'open', r) for r in file_rows),
((r[1], 'exec', r) for r in exec_rows))
for ts, event_type, data in rows:
if event_type == 'process':
r_id, r_parent, r_timestamp = data
if r_parent is not None:
parent = processes[r_parent]
binary = parent.binary
else:
parent = None
binary = None
p = Process(r_id,
parent,
r_timestamp,
False,
binary,
C_INITIAL if r_parent is None else C_FORK)
processes[r_id] = p
all_programs.append(p)
elif event_type == 'open':
r_name, r_timestamp, r_mode, r_process = data
r_name = PosixPath(r_name)
if r_mode != FILE_WDIR:
process = processes[r_process]
files.add(r_name)
edges.add((process, r_name, r_mode, None))
elif event_type == 'exec':
r_name, r_timestamp, r_process, r_argv = data
r_name = PosixPath(r_name)
process = processes[r_process]
binaries.add(r_name)
# Here we split this process in two "programs", unless the previous
# one hasn't done anything since it was created via fork()
if not all_forks and not process.acted:
process.binary = r_name
process.created = C_FORKEXEC
process.acted = True
else:
process = Process(process.pid,
process,
r_timestamp,
True, # Hides exec only once
r_name,
C_EXEC)
all_programs.append(process)
processes[r_process] = process
argv = tuple(r_argv.split('\0'))
if not argv[-1]:
argv = argv[:-1]
edges.add((process, r_name, None, argv))
process_cursor.close()
file_cursor.close()
conn.close()
# Puts files in packages
logging.info("Organizes packages...")
package_files = {}
other_files = []
for f in files:
pkg = packages.get(f)
if pkg is not None:
package_files.setdefault((pkg.name, pkg.version), []).append(f)
else:
other_files.append(f)
# Writes DOT file
with target.open('w', encoding='utf-8', newline='\n') as fp:
fp.write('digraph G {\n /* programs */\n node [shape=box];\n')
# Programs
logging.info("Writing programs...")
for program in all_programs:
fp.write(' prog%d [label="%s (%d)"];\n' % (
id(program), program.binary or "-", program.pid))
if program.parent is not None:
reason = ''
if program.created == C_FORK:
reason = "fork"
elif program.created == C_EXEC:
reason = "exec"
elif program.created == C_FORKEXEC:
reason = "fork+exec"
fp.write(' prog%d -> prog%d [label="%s"];\n' % (
id(program.parent), id(program), reason))
fp.write('\n node [shape=ellipse];\n\n /* system packages */\n')
# Files from packages
logging.info("Writing packages...")
for i, ((name, version), files) in enumerate(iteritems(package_files)):
fp.write(' subgraph cluster%d {\n label=' % i)
if version:
fp.write('"%s %s";\n' % (escape(name), escape(version)))
else:
fp.write('"%s";\n' % escape(name))
for f in files:
fp.write(' "%s";\n' % escape(unicode_(f)))
fp.write(' }\n')
fp.write('\n /* other files */\n')
# Other files
logging.info("Writing other files...")
for f in other_files:
fp.write(' "%s"\n' % escape(unicode_(f)))
fp.write('\n')
# Edges
logging.info("Connecting edges...")
for prog, f, mode, argv in edges:
if mode is None:
fp.write(' "%s" -> prog%d [color=blue, label="%s"];\n' % (
escape(unicode_(f)),
id(prog),
escape(' '.join(argv))))
elif mode & FILE_WRITE:
fp.write(' prog%d -> "%s" [color=red];\n' % (
id(prog), escape(unicode_(f))))
elif mode & FILE_READ:
fp.write(' "%s" -> prog%d [color=green];\n' % (
escape(unicode_(f)), id(prog)))
fp.write('}\n')
def graph_dot(target, runs, packages, other_files, package_map, edges,
inputs_outputs, level_pkgs, level_processes, level_other_files):
"""Writes a GraphViz DOT file from the collected information.
"""
with target.open('w', encoding='utf-8', newline='\n') as fp:
fp.write('digraph G {\n /* programs */\n'
' node [shape=box fontcolor=white '
'fillcolor=black style=filled];\n')
# Programs
logging.info("Writing programs...")
for run in runs:
run.dot(fp, level_processes)
fp.write('\n'
' node [shape=ellipse fontcolor="#131C39" '
'fillcolor="#C9D2ED"];\n')
# Packages
if level_pkgs not in (LVL_PKG_IGNORE, LVL_PKG_DROP):
logging.info("Writing packages...")
fp.write('\n /* system packages */\n')
for package in sorted(packages, key=lambda pkg: pkg.name):
package.dot(fp, level_pkgs)
fp.write('\n /* other files */\n')
# Other files
logging.info("Writing other files...")
for fi in sorted(other_files):
if fi in inputs_outputs:
fp.write(' "%(path)s" [fillcolor="#A3B4E0", '
'label="%(name)s\\n%(path)s"];\n' %
{'path': escape(unicode_(fi)),
'name': inputs_outputs[fi]})
else:
fp.write(' "%s";\n' % escape(unicode_(fi)))
fp.write('\n')
# Edges
logging.info("Connecting edges...")
done_edges = set()
for prog, fi, mode, argv in edges:
endp_prog = prog.dot_endpoint(level_processes)
if fi in package_map:
if level_pkgs == LVL_PKG_DROP:
continue
endp_file = package_map[fi].dot_endpoint(fi, level_pkgs)
e = endp_prog, endp_file, mode
if e in done_edges:
continue
else:
done_edges.add(e)
else:
endp_file = '"%s"' % escape(unicode_(fi))
if mode is None:
fp.write(' %s -> %s [style=bold, label="%s"];\n' % (
endp_file,
endp_prog,
escape(format_argv(argv))))
elif mode & FILE_WRITE:
fp.write(' %s -> %s [color="#000088"];\n' % (
endp_prog, endp_file))
elif mode & FILE_READ:
fp.write(' %s -> %s [color="#8888CC"];\n' % (
endp_file, endp_prog))
fp.write('}\n')
def dot_endpoint(self, f, level_pkgs):
if level_pkgs == LVL_PKG_PACKAGE:
return '"pkg %s"' % escape(self.name)
else:
return '"%s"' % escape(unicode_(f))
def write_cltools_module(run, dot_vistrails):
input_files = run['input_files']
output_files = run['output_files']
module_name = 'reprounzip_%s' % hash_experiment_run(run)[:7]
# Writes CLTools JSON definition
(dot_vistrails / 'CLTools').mkdir(parents=True)
cltools_module = (dot_vistrails / 'CLTools' / module_name) + '.clt'
logging.info("Writing CLTools definition %s...", cltools_module)
with cltools_module.open('w', encoding='utf-8', newline='\n') as fp:
fp.write('{\n'
' "_comment": "This file was generated by reprounzip '
'%(version)s at %(date)s",\n\n' % {
'version': version,
'date': datetime.now().isoformat()})
# python -m reprounzip.plugins.vistrails
fp.write(' "command": "%s",\n'
' "args": [\n'
' [\n'
' "constant",\n'
' "-m",\n'
' "flag",\n'
' {}\n'
' ],\n'
' [\n'
' "constant",\n'
' "reprounzip.plugins.vistrails",\n'
' "flag",\n'
' {}\n'
' ],\n' % escape(sys.executable))
# Unpacker
fp.write(' [\n'
' "input",\n'
' "unpacker",\n'
' "string",\n'
' {}\n'
' ],\n')
# Target directory
fp.write(' [\n'
' "input",\n'
' "directory",\n'
' "string",\n'
' {}\n'
' ],\n')
# Run number
fp.write(' [\n'
' "input",\n'
' "run",\n'
' "string",\n'
' {}\n'
' ],\n')
# Input files
for i, input_name in enumerate(input_files):
fp.write(' [\n'
' "input",\n'
' "input %(name)s",\n'
' "file",\n'
' {\n'
' "flag": "--input-file",\n'
' "prefix": "%(name)s:"\n'
' }\n'
' ],\n' % {'name': escape(input_name)})
# Output files
for i, output_name in enumerate(output_files):
fp.write(' [\n'
' "output",\n'
' "output %(name)s",\n'
' "file",\n'
' {\n'
' "flag": "--output-file",\n'
' "prefix": "%(name)s:"\n'
' }\n'
' ],\n' % {'name': escape(output_name)})
# Command-line
fp.write(' [\n'
' "input",\n'
' "cmdline",\n'
' "string",\n'
' {\n'
' "flag": "--cmdline"\n'
' }\n'
' ]\n'
' ],\n')
# Use "std file processing" since VisTrails <=2.1.4 has a bug without
# this (also, it's inefficient)
fp.write(' "options": {\n'
' "std_using_files": ""\n'
' },\n')
# Makes the module check for errors
fp.write(' "return_code": 0,\n')
# Enable 'stdout' port
fp.write(' "stdout": [\n'
' "stdout",\n'
' "file",\n'
' {}\n'
' ]\n'
'}\n')
return module_name