Skip to content

Callback-based filter #2076

Open
Open
@1plaintext

Description

@1plaintext

I would have thought this be a popular ask, but I didnt seem to find any discussion/solution
I am trying to absolutely minimize memory usage because the document has a huge array, the doc looks like this
{<unpredictable stuff>, "array":[{"name":<long strings>},...]}}
using a filter like this
filter["array"][0]["name"] = true;
I am still running out of memory, because, with this filter, the final product still contains all the "name"s (again, this is a very long array)

I wonder if there are other ways to more concisely filter... perhaps something like this (doesn't seem to work)
filter["array"][3-14]["name"] = true;
so I get only forth to fifteenth elements? 16 elements fits perfectly in memory vs all the elements in my case.

An alternative idea maybe a callback
filter["array"][]["name"] = callback;
so i can examine each time a "name" is found, perhaps the call back tells me the index and lets me know the "name", so I can decide to keep it, throw it away, or even stop the parsing (for speed purpose)

Now I also thought about "deserialization-in-chunks" using findUntil.. but the preceding "<unpredictable stuff>" makes it unreliable, there maybe similar named elements at different nested level etc (unless I write my own JSON parsing code, which defeats the purpose of using this library)- after all, the idea of an annotated JSON doc is so that the doc can be out of order, with additional things you don't care, etc...

Thanks for any idea.

Metadata

Metadata

Assignees

No one assigned

    Projects

    No projects

    Milestone

    No milestone

    Relationships

    None yet

    Development

    No branches or pull requests

    Issue actions