Search¶

The main page has a search box, the dictionary picker, and the results: one collapsible section per dictionary.

Search modes¶

Choose the mode in the drop-down next to the search box.

Exact and prefix work on every dictionary, indexed or not. Contains and full-text search only dictionaries that have the index; any other dictionary shows a link that adds it.

To add an index to one dictionary, click its contains or full-text switch in the dictionary panel. To add the full-text index to all dictionaries: ☰ → folder summary → Full-text for every dictionary….

Case, accents and spelling¶

All four modes ignore case and accents: corazon finds corazón, OXFORD finds Oxford.

No mode corrects spelling. Contains matches the query as a literal substring: orre finds correr, corer finds nothing.

A contains query shorter than 3 characters cannot use the index and is answered by scanning all headwords, which is slower.

Index sizes¶

For a dictionary of 40,000 entries:

The dictionary panel shows each index's size for your own dictionaries. An indexed dictionary is usually smaller than its dictionary files, because article text is stored compressed.

Full-text search¶

Full-text search matches words in the article text of dictionaries that have a full-text index. Its query has its own syntax (Full-text query syntax); in the other modes the query is literal text.

Unquoted queries¶

A query of several words, without quotes or operators, is matched in up to three stages. Each dictionary stops at the first stage that returns results.

The result header of a dictionary answered at the second or third stage shows phrase not found · words proximity or phrase not found · words anywhere.

Inflected words¶

When a single-word query returns nothing in a dictionary, DictShelf looks up the word's lemma and searches that dictionary again: knew → know, estuviera → estar. A banner names the lemma.

The retry goes only to dictionaries that returned nothing, and only with lemma data for the dictionary's language: sale → salir is never searched in an English dictionary. How the language is detected: Configuration → Lemmatization.

English lemma data is built in. To install another language:

• ☰ → folder summary → Lemmatization…, or 🔤 Lemmatization on the setup page, and tick the language. It applies from the next search. This is the only way on Android.
• In a terminal: dictshelf lemmas download ru. A running server uses it after a restart.

Untick a language to delete its lemma data.

A language's lemma data is loaded on the first search that needs it and uses 1 MB (Persian) to 90 MB (Slovak) of memory; the Lemmatization page lists each figure. At most MORPH_CACHE languages stay loaded: 2 on a desktop, 1 on Android. MORPH_CACHE = "0" turns lemmatization off.

Result order¶

Results appear in the order of the dictionary panel. The section that opens first is placed above dictionaries that have not answered yet. Open first in the dictionary panel decides which section that is: the first dictionary in your order that has a result (My order, the default), or the first to answer (Fastest).

A search of several dictionaries shows up to 5 results per dictionary; a search of one dictionary, up to 25. more… in a section header shows up to 100 results from that dictionary.

Cross-references¶

A link in an article is looked up in the same dictionary first. If that dictionary has no such headword, all dictionaries are searched, and a banner says so.

Brief and full view¶

Lingvo DSL dictionaries mark part of an article as secondary text: usually the examples, often the links to sub-entries. It is hidden when an article opens. The button at the right end of a dictionary's header shows it; its icon draws the hidden line dotted. Shown, secondary text is grey.

• The next result opens the way you left the last one. The choice is stored with your other settings, so every browser and the Android app that use the same DictShelf share it.
• A full-text match inside secondary text, or a link to a place inside it, shows it for that result only.
• Ctrl+* shows or hides it in the section you are reading (desktop).

Keys and buttons¶

Dictionary panel¶

☰ opens it.

Each dictionary row:

• ⠿ (drag) or ⏫ ▲ ▼ ⏬: reorder. Results follow this order.
• checkbox: enabled for All dictionaries searches. A disabled dictionary can still be chosen in the picker.
• name: search this dictionary only.
• switches: 🚀 index (then the index size), contains, full-text, media. Click to add or remove.
• file row (e.g. Oxford.mdx): the dictionary files, the library folder, and 🗑 Remove….
• About this dictionary: its groups and its own description.

Rescan folders re-reads the dictionary folders; use it after adding or deleting dictionary files.

Link to a search¶

The address holds the query, the mode and the dictionary selection, so a search can be bookmarked or opened from a script. Back and Forward step through searches.

http://localhost:6888/?q=serendipity&mode=prefix&dict=all

On Android, dictshelf://lookup?q=serendipity opens the same search; see the Android app.

Search from the terminal¶

dictshelf searchall phubbing                              # all dictionaries in DICT_DIR
dictshelf searchall -dict-dir ~/Dictionaries/es phubbing  # one folder
dictshelf searchall -format raw phubbing                  # the dictionaries' own HTML
dictshelf searchall -format clean phubbing                # HTML without scripts and styles

dictshelf lookup   ~/Dictionaries/Oxford.mdx phubbing
dictshelf prefix   ~/Dictionaries/Oxford.mdx phubb
dictshelf contains ~/.dmshlabs/dictshelf/db/Oxford phub
dictshelf fts      ~/.dmshlabs/dictshelf/db/Oxford "sudden fear"

contains and fts need an indexed dictionary: pass its library folder. lookup and prefix work on any dictionary file. searchall uses exact mode unless -mode says otherwise.

Full-text queries run as in the web UI. The shell removes quotes before DictShelf sees them; wrap the query in single quotes to keep double quotes:

dictshelf searchall -mode fts '"faux ami" NOT linguistics'

Command line reference