A native macOS / windows / Linux database IDE built with Swift and AppKit. Connect to PostgreSQL, MySQL, SQLite, and Redis from a single app with a fast, keyboard-driven interface.
C++
#pragma once
// Pure-function heuristics over a parsed PlanDocument. Two outputs:
// • Warnings — short, single-issue callouts shown inline (the InfoBar)
// and as `pill` badges on offending tree nodes.
// • Recommendations — actionable suggestions surfaced in the bottom
// "Recommendations" card (ANALYZE, raise work_mem, add index...).
//
// Rules are intentionally simple + transparent. We bias toward false
// negatives (no nag) over false positives — every recommendation must
// be a thing a senior DBA would also say after one glance at the plan.
#include "ExplainPlanTypes.h"
#include <string>
#include <vector>
namespace gridex::explainviz
{
enum class IssueSeverity { Info, Warn, Bad };
struct Issue
{
std::wstring nodeId; // PlanNode.id this issue points at
std::wstring title; // short summary (used in InfoBar header)
std::wstring detail; // longer body
IssueSeverity severity = IssueSeverity::Warn;
};
struct Recommendation
{
std::wstring title; // bold lead, e.g. "Create index on admissions(admit_date)"
std::wstring rationale; // single-sentence justification
std::wstring previewSql; // SQL to copy/preview when user clicks
IssueSeverity severity = IssueSeverity::Warn;
};
class ExplainPlanHeuristics
{
public:
// Walks the plan and returns the issues found. Sets level=Bad on
// any node tagged with a Bad-severity issue (mutates in place so
// the tree's heat coloring matches the InfoBar).
static std::vector<Issue> analyze(PlanDocument& doc);
// Build recommendation cards from the same pass. Caller can run
// this after analyze() so the same data is reused.
static std::vector<Recommendation> recommend(const PlanDocument& doc,
const std::vector<Issue>& issues);
};
}