Extending Persistency to Allow Associating a File to Entity and Key Pair #5476
Conversation
| bool in_current; | ||
| std::filesystem::path to_current_path() const | ||
| { | ||
| return getSaveFilePath("current", "pf-" + std::to_string(id)); |
There was a problem hiding this comment.
I would prefer that the filename prefix start with dfhack so as to clearly distinguish it as part of DFHack's cosave mechanics
| catch (std::exception&) { | ||
| // empty file? | ||
| return false; | ||
| } |
There was a problem hiding this comment.
At least log the exception reason as a diagnostic; failing silently is unkind
| for (auto& entity : json) | ||
| { | ||
| if (entity.isMember("e")) | ||
| { | ||
| auto pair = file_storage.try_emplace(entity["e"].asInt()); | ||
| if (entity.isMember("arr")) | ||
| { | ||
| auto arr = entity["arr"]; | ||
| if (arr.isArray()) | ||
| { | ||
| for (auto& k_v : arr) | ||
| { | ||
| if (k_v.isMember("k") && k_v.isMember("v")) | ||
| { | ||
| const size_t this_idx = k_v["v"].asUInt64(); | ||
| max_idx = std::max(max_idx, this_idx); | ||
|
|
||
| pair.first->second[k_v["k"].asString()] = FileEntry{this_idx, false}; | ||
| //file_storage[e][k] = v | ||
| } | ||
| } | ||
| } | ||
| } | ||
| } | ||
| } | ||
| } |
There was a problem hiding this comment.
i don't like the deep nesting here and would suggest using std::ranges::filter to prefilter the iterators instead
| int entity_id = Persistence::WORLD_ENTITY_ID; | ||
| if (fname != "dfhack-world.dat" && !get_entity_id(fname.string(), entity_id)) | ||
| continue; | ||
| if (fname == "dfhack-extra-files") |
There was a problem hiding this comment.
using file names with no extension is rude to windows platform users
|
|
||
| std::filesystem::path Persistence::getFile(int entity_id, const std::string& key, bool* added, bool just_for_reading) | ||
| { | ||
| if (!is_good_entity_id(entity_id) || key.empty() || !Core::getInstance().isWorldLoaded()) |
There was a problem hiding this comment.
this exact test appears at least three times on lines 580, 600, 610, and in part in 645 and 673 (and possibly other places). should this be abstracted into a utility function to avoid repeating yourself?
I propose a set of functions to
DFHack::Persistencyto allow managing whole files, following the same entity ID and key logic as the pre-existing functions.The API was designed around managing the possible paths for these files rather than the files themselves; it would be up to the callers to interact with them according to the demands of their particular use case. Every (entity ID, key) pair will be associated with a unique file ID, corresponding to a file with a name of the form
dfhack-pf-<ID>.dat. The association between (entity ID, key) pairs and file IDs is saved as JSON indfhack-extra-files.dat.Files that are requested will be copied over to the
currentfolder unless specifically requested as read only through the API (in which case modifying their contents leads to these changes not being reflected if the user does not save to the same folder).No attempt is made at this stage to offer any particular form of synchronization or memory management of the files themselves. Operations that add to or query the list of available files are guarded by
CoreSuspender, but this ultimately only protects the list of file IDs.No provisions are made for associating more than one file to an (entity ID, key) pair.
Note: I surmised from perusing previous discussions that it was not yet consensual among the main developers whether DFHack should offer the ability to request an entire file for persistency. However, for my current project of overhauling the dwarf logic system, serialization/deserialization could be important for performance if the user configures many objects to be handled by my system, so a more optimized way of handling that seems desirable. My apologies if I overstep any boundaries by proposing this.