Connect machinery to ingest previous reports

This is work towards #946

This commit connects the command line argument --with-context or
-ctx to the inventory machinery for the --live option.

- We introduce a consumer entrypoint for the format plugin json.
  We make the decision here to keep the formats of the input and
  output reports to be consistent. In the JSON case, the report
  provided as -ctx context must be of JSON format. We do this to
  minimize the overhead of maintaining a set of formats for input
  reports as well as output reports and thus help with code reuse.
  For the JSON format, we use "jsonc".
- In the --live run, we modify the actual format selector string to
  add the "c" at the end of it, thus allowing us to call the
  consumer plugin rather than the generator plugin. The reporting
  will stay the same.
- We also added some fixes to the Consumer abstract base class to
  implement "consume_layer" rather than just "consume", and fixed
  the layer index increment in the JSON Consumer class.

Signed-off-by: Nisha K <[email protected]>

setup.cfg

@@ -47,6 +47,7 @@ tern.formats =
     spdxtagvalue = tern.formats.spdx.spdxtagvalue.generator:SpdxTagValue
     spdxjson = tern.formats.spdx.spdxjson.generator:SpdxJSON
     json = tern.formats.json.generator:JSON
+    jsonc = tern.formats.json.consumer:JSON
     yaml = tern.formats.yaml.generator:YAML
     html = tern.formats.html.generator:HTML
 tern.extensions =

tern/analyze/default/live/run.py

@@ -12,6 +12,8 @@
 """
 import logging
 import os
+from stevedore import driver
+from stevedore.exception import NoMatches
 
 from tern.utils import constants
 from tern.utils import rootfs
@@ -86,6 +88,27 @@ def fill_packages(layer, prereqs):
         com.remove_duplicate_layer_files(layer)
 
 
+def get_context_layers(reports, format_string):
+    """Given a list of reports and the format string, get corresponding layer
+    objects for each of the reports. We will load the required module and
+    run the consume_layer function to return the list of layers"""
+    # we want to maintain consistency between the input report formats and
+    # the output formats by redirecting the command line argument to the
+    # correct consumer entrypoint. Hence we maintain a known pattern for
+    # entrypoint strings which is simply that the corresponding consumer for
+    # a given generator is the generator's entrypoint string + 'c'
+    consumer_format_string = format_string + 'c'
+    try:
+        mgr = driver.DriverManager(
+            namespace='tern.formats',
+            name=consumer_format_string,
+            invoke_on_load=True,
+        )
+        return mgr.driver.consume_layer(reports)
+    except NoMatches:
+        pass
+
+
 def execute_live(args):
     """Execute inventory at container build time
     We assume a mounted working directory is ready to inventory"""
@@ -105,5 +128,15 @@ def execute_live(args):
     prereqs.host_shell = host.check_shell()
     # collect metadata into the layer object
     fill_packages(layer, prereqs)
+    # resolve unique packages for this run with reports from previous runs
+    if args.with_context:
+        # get a list of previous layers based on plugin type
+        context_layers = get_context_layers(
+            args.with_context, args.report_format)
+        # resolve the packages for each of the layers
+        context_layers.append(layer)
+        master_list = []
+        for l in context_layers:
+            dcom.update_master_list(master_list, l)
     # report out the packages
     report.report_layer(layer, args)

tern/formats/consumer.py

@@ -10,8 +10,8 @@
 class Consume(metaclass=ABCMeta):
     """Base class for report consuming plugins"""
     @abstractmethod
-    def consume(self, reports):
-        """Ingest the contents of the list of files into an Image object or
-        ImageLayer object according to the plugin type. Each plugin is
+    def consume_layer(self, reports):
+        """Ingest the contents of the list of files into a list of ImageLayer
+        objects according to the plugin type. Each plugin is
         responsible for implementing the reading and and assimilation of the
         report metadata"""

tern/formats/json/consumer.py

@@ -70,5 +70,5 @@ def consume_layer(self, reports):
             layer = create_image_layer(report)
             layer.layer_index = layer_count
             layer_list.append(layer)
-            layer += 1
+            layer_count += 1
         return layer_list