Method ReadMatchesAsync

Namespace

MailFathom.Application.Emails.Search

Assembly

MailFathom.Application.dll

ReadMatchesAsync(MailboxEmailSelection, EmailSearchQueryText, EmailSearchSnippetBounds, IReadOnlyList<RankedEmailCandidate>, CancellationToken)

Reads the summaries and snippets of an already ranked window, in the order it was ranked.

Task<IReadOnlyList<EmailSearchMatch>> ReadMatchesAsync(MailboxEmailSelection selection, EmailSearchQueryText queryText, EmailSearchSnippetBounds snippetBounds, IReadOnlyList<RankedEmailCandidate> rankedCandidates, CancellationToken cancellationToken)

Parameters

selection MailboxEmailSelection

The same filters the candidates were ranked under.

queryText EmailSearchQueryText

The validated free text the snippets are cut around.

snippetBounds EmailSearchSnippetBounds

How many extracts one result may carry, and how long each may be.

rankedCandidates IReadOnlyList<RankedEmailCandidate>

The window to read, best first, each carrying the score the result publishes.

cancellationToken CancellationToken

Propagates caller cancellation.

Returns

Task<IReadOnlyList<EmailSearchMatch>>

One match per candidate that is still eligible, in the candidates' order.

Remarks

The filters are re-applied rather than trusted from the ranking, because the two statements read different snapshots: a run committing between them can leave a message the ranking admitted outside the request's scope, or delete it outright. Such a candidate is absent from the result rather than published, so a window never contradicts the request that produced it.

Snippets are cut around queryText whether the candidate was ranked lexically or by similarity. A semantically ranked message that carries none of the query's words therefore carries no snippets, which is the honest answer: there is no extract of it that shows why it matched, and returning its opening words would present the start of a body as though it were a match.

Exceptions

ArgumentNullException

Thrown when any reference argument is null.