Search Grammar

Type like you think. The search understands.

JinDesk search bar accepts a progressive grammar — beginners type words, experts type expressions, both get results.

---

Six Levels of Syntax

Level 1 — Just type

The simplest case: type any words. Each word is matched as a word-start — the term must appear at the start of a word in one of the visible columns (a word ends at a space, underscore, hyphen, or dot). Multiple words become AND filters. Double-quote an exact phrase.

paracetamol                  → any word starts with "paracetamol"
hip replacement              → AND of two word-start matches
"Scalpel Blade"              → the phrase at a word boundary

If you want “contains anywhere” (not just word-start), use ~ — see Level 3.

Level 2 — Operators

Comparisons on numeric and date columns. Ranges with ...

>1000                        → any numeric column > 1000
<50                          → any numeric column < 50
100..500                     → any numeric column between 100 and 500
2025-01..2025-06             → any date column in this range
>=high                       → ordinal: low < medium < high

Ranges also work for dates at month granularity (2025-01..2025-06), day granularity (2025-04-01..2025-04-30), or mixed.

Level 3 — Match operators

Three prefixes control how a term matches:

Prefix	Meaning
(none)	word-starts-with (default)
=	exact match
~	regex — also how you do “contains anywhere”

Applies both to bare terms and to field-qualified ones (next level):

ortho                        → any word starts with "ortho"
=ortho                       → exactly "ortho"
~ortho                       → regex / contains "ortho" anywhere
~^ortho                      → regex anchored to column start

Level 4 — Field-specific filters

Scope a match to a single column with !field:value. The three prefixes above apply here too:

!amount:>1000                → amount column > 1000
!date:2025-01..2025-06       → date in range
!material_id:630664          → word-start on material_id
!material_id:=630664         → exact match
!name:~hip|knee              → regex on name column
!status:null                 → IS NULL
-!status:null                → IS NOT NULL (negation; there is no separate "not_null" syntax)
!is_active:true              → boolean

Field filters compose naturally — use them for one column, leave plain words for the rest.

Level 5 — Logic

Combine terms with OR, exclude with -, group with parentheses.

hip OR knee                  → either term matches
hip | knee                   → shorthand for OR
-consumable                  → NOT: exclude rows matching "consumable"
!cat:ortho -!s:pending       → ortho category, NOT pending status
hip (ortho | cardiac)        → hip AND (ortho OR cardiac)

Whitespace between terms is AND. OR / | between terms is OR. - in front of a term (or -!field:value) negates it.

Level 6 — Smart date tokens

Natural dates that expand to real date ranges against any date column. this-* tokens use the full calendar period (not “up to today”) so the result stays the same whether you ask on the 1st or the 30th of the month.

yesterday                    → yesterday's date (one day)
today                        → today (one day)
tomorrow                     → tomorrow (one day)

last-7d                      → last 7 days, ending today
last-30d                     → last 30 days, ending today
next-7d                      → next 7 days, starting today
next-30d                     → next 30 days, starting today
last-Nd / next-Nd            → any N works

last-week                    → full previous week
this-week                    → full current week
next-week                    → full following week

last-month                   → full previous calendar month
this-month                   → full current calendar month
next-month                   → full following calendar month

last-year                    → full previous calendar year
this-year                    → full current calendar year
next-year                    → full following calendar year

last-week-3                  → the week that was 3 weeks ago
next-week-2                  → the week 2 weeks from now
last-month-12                → the month 12 months ago
next-year-5                  → the year 5 years from now

The stepped forms (last-week-N, last-month-N, last-year-N and their next- counterparts) give you any offset without reaching for literal dates. Three handy identities: last-week-1 ≡ last-week; last-week-0 ≡ this-week; a “trailing 12 months” window is simply last-month-12..last-month-1.

The next-* family lands squarely on future dates — useful for columns that store intent (material expiration, contract end, scheduled admission, warranty boundary).

Week boundaries follow display.search.week_start (default: Monday, per ISO 8601 / DACH convention). Set it to sunday in config.yml if you want calendar-week semantics instead.

Smart date tokens also compose inside ranges — either or both sides of .. can be a token:

last-year..this-year         → start of last year to end of this year
today..next-week             → today through end of next week
this-week..today             → Monday of this week through today
last-month..2026-06-30       → full previous month through that literal date

The left side of the range uses the token’s start; the right side uses its end. Everything else the range accepts (ISO dates, pure numbers) still works on either side.

Range bounds are order-agnostic. 500..100 means the same as 100..500; next-year..last-year means the same as last-year..next-year. The parser / SQL layer always binds the interval as (min, max), so misordered bounds never silently return zero rows.

Period + day truncation (the useful special case). When a period token (this-month, this-year, etc.) is combined with a single-day token (today), the range uses the period’s inner edge, giving you the common month-to-date / year-to-date / week-to-date idioms:

this-month..today            → 1st of this month through today (MTD)
this-year..today             → Jan 1 through today (YTD)
this-week..today             → Monday of this week through today (WTD)
today..this-week             → today through Sunday of this week

Under the hood: the left bound contributes its .start, the right bound contributes its .end. If that produces a reversed pair (e.g. next-year..last-year), the parser falls back to the union of both spans — so order-agnostic symmetry still holds for reversed inputs.

Open ranges. Leave one side blank to make the range unbounded on that side — X.. means “from X onwards,” ..Y means “up to and including Y.” Works for numbers, ISO dates, and smart date tokens:

100..                        → any numeric column >= 100
..500                        → any numeric column <= 500
today..                      → from today onwards  (today or later)
..next-month                 → up to the end of next month
last-year..                  → from Jan 1 of last year onwards
!expires:today..next-month   → anything expiring between now and end of next month
!shipped:..yesterday         → everything shipped before today

Open ranges are often what you actually want for expiration, renewal, or shipment-date filters. last-year..this-year gives you a fixed 2-year window; last-year.. keeps rolling forward forever as time passes.

---

Parsing Rules

1. Whitespace splits terms (AND by default).
2. Quoted strings are single terms: "hip replacement".
3. OR or | between terms creates an OR group.
4. - prefix negates: -term, -!field:value.
5. !field:value is a field filter.
6. >, <, >=, <= before a number = comparison.
7. N..M = range (numeric or date).
8. Smart tokens (today, this-month, last-7d) expand to date ranges.
9. Parentheses group OR expressions: (a | b) c.
10. Backslash escapes let you embed spaces, quotes, or parens in an unquoted value: !name:hip\ replacement, !note:\"quoted\". \d, \., \[ etc. pass through unchanged so regex patterns work.

---

Finding help while you type

Every search bar has an ⓘ button on the right. Click it for a compact quick-reference of the grammar — Basic / Operators / Field filters / Numbers / Logic / Dates — with a single literal example per row. Press ESC or click outside to close.

The URL updates when you press Enter or when the input loses focus. Typing itself is purely local — no re-render mid-keystroke.

---

Worked Examples

Filter 68,000 cases down to what you care about

“High-severity materials over CHF 1,000 from last year”

!severity:=high !price:>1000 !date:2025-01..2025-12

Three field filters composed with implicit AND. One input, one query.

Find all ortho or cardiac cases that are not pending

!cat:=ortho | !cat:=cardiac -!status:=pending

OR at the top level, then a negated exact field filter.

Find expensive recent findings

CHF >10000 last-30d

Shorthand + smart token. Works in finding views where CHF maps to money_at_risk.

Pin a specific material

=635465

Exact match across all visible columns — any column whose value equals 635465 matches.

Configuring the word-boundary set

Per tenant, in config.yml:

display:
  search:
    case_sensitive: false          # default
    word_separators: " _-."        # chars that start a new "word"
    week_start: monday             # `monday` (default) or `sunday`

Tighten the separator set for stricter word-start matching; widen it to find substrings inside more identifier shapes. week_start changes what last-week / this-week / next-week consider a “week.”

---

Related

• JinDesk Guide — where search lives in the UI