Optimize Dictionary Import Memory Use

[Casheeew]

The massive memory usage by loading term files all into memory may cause Chrome to be slower and sometimes have issues such as #1420 .

This PR instead achieves a constant memory usage via using a streaming pattern:
This PR uses zip.js' getData() stream to pipe into a custom JSON parser, and emits onBatch events whenever there are maxTransactionLength entries. This effectively reduces the memory usage by half even when the term files are small (at 1000 terms / term bank), and caps the memory usage at about 40 MB.

works flawlessly even when combining pixiv full into one humongous term bank

Based on my comprehensive testing, there is negligible difference in import time. this is slightly faster due to less GC

Maybe a partial fix to the many "Unknown error" issues... #536

[Casheeew]

@codex review

[chatgpt-codex-connector]

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: b1694dafcd

ℹ️ About Codex in GitHub

Codex has been enabled to automatically review pull requests in this repo. Reviews are triggered when you

• Open a pull request for review
• Mark a draft as ready
• Comment "@codex review".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

When you sign up for Codex through ChatGPT, Codex can also answer questions or update the PR, like "@codex address that feedback".

[Casheeew]

@codex review

[chatgpt-codex-connector]

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: 00ca6997b5

ℹ️ About Codex in GitHub

Codex has been enabled to automatically review pull requests in this repo. Reviews are triggered when you

• Open a pull request for review
• Mark a draft as ready
• Comment "@codex review".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

When you sign up for Codex through ChatGPT, Codex can also answer questions or update the PR, like "@codex address that feedback".

[chatgpt-codex-connector]

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: 1981762c53

ℹ️ About Codex in GitHub

Codex has been enabled to automatically review pull requests in this repo. Reviews are triggered when you

• Open a pull request for review
• Mark a draft as ready
• Comment "@codex review".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

When you sign up for Codex through ChatGPT, Codex can also answer questions or update the PR, like "@codex address that feedback".

[Casheeew]

Should be good to go. We don't have to cover all edge cases since realistically dictionaries should be for the large part generated by code, so there is little reason to check every malformed JSON. The main check is the AJV layer and if the dictionary passes, it should not bork the user in any way with malformed data.

[Casheeew]

@codex review

[chatgpt-codex-connector]

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: fdf7851b2b

ℹ️ About Codex in GitHub

Codex has been enabled to automatically review pull requests in this repo. Reviews are triggered when you

• Open a pull request for review
• Mark a draft as ready
• Comment "@codex review".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

When you sign up for Codex through ChatGPT, Codex can also answer questions or update the PR, like "@codex address that feedback".

[Kuuuube]

Based on my comprehensive testing, there is negligible difference in import time. this is slightly faster due to less GC

Can you give a few timed tests of imports and the peak memory usage youre seeing before and after?

[Casheeew]

Okay, I tested again. For dictionaries which term files have 1000 entries the performance is comparable (both versions are loading 1000 entries into memory, I probably got the impression that it was halved from a wrong tab..)
import time is quite high variance (few seconds for EN-KRDICT, >10s for Pixiv)

PixivFull before (210.54s)
validation step

import step

PixivFull after (203.00s)
validation step

import step

for EN-KRDICT where each term file has a different number of entries, the memory is effectively halved:

Before (12.41s):

After (13.20s):

bonus: Pixiv with 1 term bank (700k+ entries)

Before: ∞ seconds
import error
After: 205.44s
validation step

import step

basically equivalent to when split into 1000 entry term banks