TODO: update JSON and other changes

TODO

@@ -11,19 +11,10 @@ Before Release:
 Next:
 * output capture
 * QPDFPagesTree -- avoid ever flattening the pages tree.
+* JSON v2 fixes
 
 Pending changes:
 
-* Think about whether additional JSON use cases would be served by
-  having qpdf-v2 contain things other than "objects" or making qpdf
-  --json include everything that --json-output includes. Right now, if
-  you wanted to do something in json objects based on page
-  information, you'd have to run qpdf --json and also qpdf
-  --json-output separately. Also "qpdf-v2" doesn't follow the naming
-  convention. See pinned email from m-holger with subject "qpdf
-  json.rst" from June 5
-  (04ad60e5-3274-4a9e-abde-3de97640d370@www.fastmail.com)
-* Good C API for json v2
 * Check about runpath in the linux-bin distribution. I think the
   appimage build specifically is setting the runpath, which is
   actually desirable in this case. Make sure to understand and
@@ -43,14 +34,10 @@ Pending changes:
   reveal additional details, --show-encryption could potentially retry
   with this option if the first time doesn't work. Then, with the file
   open, we can read the encryption dictionary normally.
-* Nice to have:
-  * In libtests, separate executables that need the object library
+* In libtests, separate executables that need the object library
   from those that strictly use public API. Move as many of the test
   drivers from the qpdf directory into the latter category as long
   as doing so isn't too troublesome from a coverage standpoint.
-  * Rework tests so that nothing is written into the source directory.
-    Ideally then the entire build could be done with a read-only
-    source tree.
 * Consider adding fuzzer code for JSON
 
 Soon: Break ground on "Document-level work"
@@ -128,6 +115,78 @@ sure /Count and /Parent are correct.
 refs/attic/QPDFPagesTree-old -- original, abandoned branch -- clean up
 when done.
 
+
+JSON v2 fixes
+=============
+
+* Get rid of separate format for --json and --json-output. Instead,
+  --json-output can just require an outfile and change some defaults
+  like which keys are present and json-stream-data. This makes it
+  easier to support use cases like being able to use information in
+  other top-level keys ("pages", "attachments", etc.) to drive
+  modifications made to objects without having to run qpdf twice. I
+  think --json-output should make the default key be only "qpdf" and
+  the default json-stream-data mode be inline, but make it so you can
+  use --json-stream-data and --json-stream-prefix with --json and
+  --json-keys with --json-output. These would be exactly the same:
+
+  --json-output --json-keys=all -
+  --json --json-stream-data=inline
+
+  And these:
+
+  --json-output -
+  --json --json-stream-data=inline --json-key=qpdf
+
+* Change the name of the "qpdf-v2" key to "qpdf". Use that in place of
+  "objects" and change its content to a two-element array whose first
+  element is metadata required (or useful) for parsing and whose
+  second element contains the actual data. Use of an array is the only
+  way to ensure that the metadata is guaranteed to be parsed before we
+  start parsing the objects. Example:
+
+  {
+    "qpdf": [
+      {
+        "jsonversion": 2,
+        "repairpagestree": false,
+        "maxobjectid": 10
+      },
+      {
+        "pdfversion": "1.3",
+        "objects": {
+          ...
+        }
+      }
+    ]
+  }
+
+  This implies a few things:
+
+  * QPDF::writeJSON will have to take an argument indicating whether
+    additional keys are being written which determines whether it
+    outputs the outer braces or not.
+
+  * This changes the policy about additional extra keys. Have a
+    guarantee that qpdf will never add a key whose name is or starts
+    with "xdata". We still have to ignore unknown keys for future
+    compatibility, but at least this gives people a namespace they can
+    know will never conflict with future keys.
+
+  * Change schema validation so that if the schema contains an array
+    with more than one element, the output has to have an array with
+    the same number of elements whose individual elements are
+    validated according to the regular rules.
+
+* Support json v2 in the C API. At a minimum, write_json,
+  create_from_json, and update_from_json need to be there and should
+  take the same kinds of functions as the C API for logger.
+
+* Address json.rst comment from m-holger: "The discussion of stream
+  objects is very wordy. Would a table similar to the style of the PDF
+  spec be easier to use?"
+
+
 Possible future JSON enhancements
 =================================
 
@@ -439,7 +498,9 @@ I find it useful to make reference to them in this list.
 
  * Look at https://bestpractices.coreinfrastructure.org/en
 
- * Get rid of remaining assert() calls from non-test code.
+ * Rework tests so that nothing is written into the source directory.
+   Ideally then the entire build could be done with a read-only
+   source tree.
 
  * Large file tests fail with linux32 before and after cmake. This was
    first noticed after 10.6.3. I don't think it's worth fixing.