BUG/ENH: make attachements compatible with kids, and allow list in RF #2197
Conversation
pubpub-zz
changed the title
Codecov ReportAttention:
Additional details and impacted files@@ Coverage Diff @@
## main #2197 +/- ##
==========================================
- Coverage 94.45% 92.53% -1.93%
==========================================
Files 43 43
Lines 7650 7930 +280
Branches 1511 1576 +65
==========================================
+ Hits 7226 7338 +112
- Misses 262 393 +131
- Partials 162 199 +37
☔ View full report in Codecov by Sentry. |
|
@MartinThoma |
|
@MartinThoma therefore attachements should return a dictionary where keys are the names in the dictionnary the point to output the return within a list is here to not change the interface. based on the better understanding of the spec, a list can not have more than on entry. |
|
I'm sorry, I don't understand the dilemma. What are the options we have?
I don't understand. Can you give me a pseudo-code example? |
As the key in the name tree are uniques, the attachements can only return one file specification stored in my dilemma is: we will get the following interface: class AttachmentBytes(bytes):
...
class PdfReader: # the same for PdfWriter
...
@property
def attachments(self) -> Mapping[str, Union[AttachmentBytes, Dict[str, AttachmentBytes]]]:
... |
|
The current signature is def attachments(self) -> Mapping[str, List[bytes]]What is Changing the signature to |
|
@MartinThoma |
|
@MartinThoma |
|
@MartinThoma |
|
Sorry, I forgot about this 🙈 A bit of notes for myself: RF is short for "Related Files" and "EF" is short for "Embedded File"; see Table 44 – Entries in a file specification dictionary. It is an optional entry with a dictionary:
This PR attempts to solve two issues:
I'd prefer to have a PR + tests which deal with #2087 first before we fix #2090 or do bigger refactorings. First dealing with #2087 would make the current PR smaller, right? |
| @property | ||
| def attachments_names(self) -> List[str]: | ||
| """ | ||
| Returns: | ||
| dictionary of filename -> Union[bytestring or List[ByteString]] | ||
| if the filename exists multiple times a List of the different version will be provided | ||
| List of names | ||
| """ | ||
| catalog = cast(DictionaryObject, self.trailer["/Root"]) | ||
| # From the catalog get the embedded file names | ||
| try: | ||
| filenames = cast( | ||
| ArrayObject, | ||
| cast( | ||
| DictionaryObject, | ||
| cast(DictionaryObject, catalog["/Names"])["/EmbeddedFiles"], | ||
| )["/Names"], | ||
| ) | ||
| except KeyError: | ||
| return {} | ||
| attachments: Dict[str, Union[bytes, List[bytes]]] = {} | ||
| # Loop through attachments | ||
| for i in range(len(filenames)): | ||
| f = filenames[i] | ||
| if isinstance(f, str): | ||
| if filename is not None and f != filename: | ||
| continue | ||
| name = f | ||
| f_dict = filenames[i + 1].get_object() | ||
| f_data = f_dict["/EF"]["/F"].get_data() | ||
| if name in attachments: | ||
| if not isinstance(attachments[name], list): | ||
| attachments[name] = [attachments[name]] # type:ignore | ||
| attachments[name].append(f_data) # type:ignore | ||
| else: | ||
| attachments[name] = f_data | ||
| return attachments | ||
| return self.attachments.keys() |
There was a problem hiding this comment.
I'd prefer not to have the new attachments_names property. I think users and pypdf itself should call self.attachments.keys() directly.
| F = "/F" # A file specification string of the file as described in Section 3.10.1 | ||
| UF = "/UF" # A unicode string of the file as described in Section 3.10.1 | ||
| EF = "/EF" # dictionary, containing a subset of the keys F , UF , DOS , Mac , and Unix | ||
| RF = "/RF" # dictionary, containing arrays of /EmbeddedFile | ||
| DESC = "/Desc" # description of the file as de |
There was a problem hiding this comment.
Those could be in an individual PR - that would be easy to review/quick to merge :-)
closes #2087
closes #2090
add also compatibility with RF (adding list)
still in progress