Bloated dictionaries come with many issues:
- From the user side they are slow to import, and contain many noisy, useless definitions/redirections.
- They take a lot of memory that can OOM the
release.rs command.
- And that forces a MAX_NUM_THREADS limit (2) that could be higher, had we used less memory.
While the OOM can be tackled in other ways, the user experience remains unpleasant.
The reasons of the bloat can be multiple:
- en-en has very heavy lemmas
- fr-fr and de-de have faulty forms (they contain articles etc.). This was somewhat fixed for de-de.
- fi-en has a lot of (non faulty, the language is like that) forms
- el-el edition love person names, they account for half the entries.
- etc.
It would be nice to trim, at least the main offenders, to keep them in line with the group.
Rank Size Dictionary
--------------------------------------------------------------------------------
1 111.0 MB main/en -> en
2 80.6 MB main/fr -> fr
3 79.8 MB main/ru -> ru
4 59.3 MB main/de -> de
5 41.3 MB main/fi -> en
6 37.5 MB main/el -> el
7 35.7 MB main/nl -> nl
8 31.8 MB main/es -> es
9 26.0 MB main/la -> en
10 25.4 MB main/ru -> en
Update after 09/04/2026
Rank Size Dictionary
--------------------------------------------------------------------------------
1 94.6 MB main/en -> en
2 78.7 MB main/ru -> ru
3 78.5 MB main/fr -> fr
4 52.0 MB main/de -> de
5 41.0 MB main/fi -> en
6 35.4 MB main/nl -> nl
7 34.9 MB main/el -> el
8 31.5 MB main/es -> es
9 24.6 MB main/ru -> en
10 24.3 MB main/la -> en
Update after 24/04/2026
Rank Size Dictionary
--------------------------------------------------------------------------------
1 95.4 MB main/en -> en
2 78.3 MB main/ru -> ru
3 71.4 MB main/fr -> fr
4 50.3 MB main/de -> de
5 41.1 MB main/fi -> en
6 35.7 MB main/el -> el
7 26.5 MB main/nl -> nl
8 24.6 MB main/ru -> en
9 24.2 MB main/es -> es
10 24.1 MB main/la -> en
Track:
Bloated dictionaries come with many issues:
release.rscommand.While the OOM can be tackled in other ways, the user experience remains unpleasant.
The reasons of the bloat can be multiple:
It would be nice to trim, at least the main offenders, to keep them in line with the group.
Update after 09/04/2026
Update after 24/04/2026
Track: