User Journey: Find a convention by its terms

"I know the terms: AED, annual, ACT/365. Which document in this kind states them, or is there none?"

The journey is done when the person holds the matching document, or holds the fact that none matches and can go on to author one. A person who ends without a document must know that the search read every row, and not only the first page.

The party holds 320 swap documents that state 201 distinct term sets. A person cannot answer this question by reading the list, and four other journeys depend on the answer. See Related journeys.

The actor is any person who reads or writes conventions: market risk, a trading desk, operations or a tenant administrator. The same activity serves all four. The visible rows differ by permission and by row-level policy, and the screen is the same.

The trigger is a set of terms the person holds and a document they need. It comes from a trade that states the terms, a counterparty agreement, a run whose terms are in doubt, or the wish to author a document without writing a duplicate. The journey starts on one kind and ends on a matched document or on a statement that none exists.

The journey keeps no server state. This journey reads. It writes nothing.

1. Choose the kind. The person picks the kind that serves the instrument family. The system shows how many documents the party holds in the kind and how many distinct term sets they state. For a tenant kind it states that the documents belong to every party.
2. State the terms. The system draws one row for each term column the kind carries. The person sets the value they hold. A term left as any matches every value. A term that is a pick is chosen from the list that owns it, so a code that no list holds cannot be typed. Empty and any are different: empty matches only a row that states nothing for the term.
3. Read the matches. The system lists the documents whose terms equal the stated ones. Each row shows the key and the terms the person left as any, so the person sees how the matches differ. The system says how many rows it read, because a search that read a page is not a search of the kind.
4. Read what a match is. The person opens one match. The system shows its terms in words, its version, where it came from, which sets name it and whether the party's runs read it.
5. Find that none matches. When no document states the terms, the screen says so in one sentence, with the number of rows read, and offers two ways on: author a document with these terms, or change one term and search again.
6. Leave, or go on. The person goes to Author a convention for a kind with the terms carried over, or to another journey with the matched key.

One screen of four panels: the kind, the terms, the matches and the empty result. The terms panel is the one the journey is about. The search is a set of equals on the columns the kind carries, and the screen draws the columns of the chosen kind only.

A clickable prototype draws this journey: prototype, with its design note find-a-convention-by-its-terms. It opens with no build and reads mock data only. The page marks each server gap this journey names.

1. The list cannot be searched by a term. The generated filter takes id_one_of and nothing else. A match on terms is a read of every page followed by a comparison in the browser. The screen can do this, and on a kind of 320 documents it is a real cost. The candidate is a filter on the term columns of each kind, for example refdata.v1.<kind>_conventions.list with a terms_equal argument. Until it exists the screen reads every page itself and says how many rows it read.
2. The search differs for each kind. The 26 tables have different columns, so there is no shared filter. A generated filter is the only shape that does not claim a column a table lacks.
3. No term-set count comes from the server. The count of 201 distinct term sets is a number the screen must compute. It is a derived fact, and it is read once for each search.
4. A match cannot say whether a run reads it. No default marker exists, and a configuration holds no rows. The screen says so and does not leave the column empty.
5. A tenant kind is searched across every party. currency_pair, ibor_index, overnight_index and tenor carry no party_id. The screen states the scope.
6. The page route has no filter, no order and no search. GET /api/conventions/<family> sends an empty filter and no order. The journey adds a terms argument, for example GET /api/conventions/<family>?terms…=.

• Author a convention for a kind — the next step when no document matches
• Change the convention a party uses for one kind — the next step when a match needs a correction
• Retire a convention — find the twin of a document before it is closed
• Name a convention set — find the document a set names for a kind
• See the conventions a party uses — read the kind as a whole

• Conventions journeys — the group this journey belongs to
• User Journeys — the index

• User Journey Extraction Standard — the checklist this journey is written against
• Document type: user_journey — the sections this page fills