File Search Engine · Analysis

01 · Frame the problem

Understand the central challenge

Design a reusable file search library over a supplied directory provider. Queries combine name, type and size predicates with AND/OR/NOT. Traversal needs cancellation, budgets and explicit error behavior.

Traversal and matching have different responsibilities and failure modes. Analyse predicate semantics, inaccessible entries and the point where work limits make a result incomplete rather than an authoritative empty answer.

02 · Find the design pressure

Start simple, then identify the gap

Begin with pure predicates over normalized entries and an iterative provider traversal. Mixing callbacks with directory recursion makes composition and error handling fragile. Bound depth/work/results and return partial coverage explicitly when traversal cannot finish.

Build a pure predicate expression tree over normalized entries and an iterative traversal engine over a provider interface. Explicit budgets and partial-result status bound work and make failures visible. Define Entry metadata and pure predicates. And/Or short-circuit, Not negates a predicate result, and leaf predicates specify their supported entry types and units. Use an explicit stack or iterator for traversal rather than an unbounded recursive stack. A provider lists child metadata; the search layer never mutates storage. Before each visit, check cancellation and depth/visit/result budgets. This exercise skips link traversal; providers supply stable identities and repeated directory identities are skipped defensively. Permission/provider errors are collected and traversal continues under a skip-and-report policy. Return results plus COMPLETE, PARTIAL, CANCELLED or BUDGET_EXCEEDED status; ordering follows a documented provider/sort policy.

03 · Define success

Requirements to preserve

Represent queries as composable predicates.
Separate traversal from metadata/storage access.
Bound depth, results and visited entries.
Define link/cycle and permission-error policies.

04 · Identify the correctness boundary

Invariants, representation and mutation boundaries

Compile predicates independently from traversal so name, size and boolean combinations can be tested as pure rules. Iterative traversal avoids call-stack dependence, while explicit entry/depth/result budgets prevent unbounded work. Provider errors and exhausted budgets require a partial-result status distinct from a complete empty result.

05 · Make it concrete

Reason through a worked example

Search for large files under a directory with an unreadable child. Matching files from readable branches may be returned with the access failure recorded. Reaching the result budget marks truncation. The caller must not infer that omitted branches contained no matching files.