Opt-in behavioural knobs that the ToSql migration-plan renderer honours. More...

#include <MigrationPlan.hpp>

Classes
struct	ColumnKey

struct	TableKey
	Tables for which the lookup has already been attempted, so we don't re-query for every INSERT against the same table. More...

Public Types
enum class	WidthUnit : uint8_t { Characters , Bytes }
	How a column's declared width is counted. The truncation layer must match the server's interpretation: a VARCHAR(100) on MSSQL holds 100 bytes, but an NVARCHAR(100) holds 100 characters. Truncating to 100 characters when the server budget is 100 bytes still overflows on multi-byte source data. More...

Public Attributes
bool	lupTruncate = false

std::map< ColumnKey, ColumnWidth >	columnWidths
	Cache of declared character widths, populated lazily as plan elements render.

std::function< void(MigrationRenderContext &, std::string_view, std::string_view)>	widthLookup
	Optional fallback that fetches column widths for a `(schema, table)` from the live database when the cache has no entry for it.

std::set< TableKey >	lookupAttempted
	Tables for which `widthLookup` has already been called; prevents repeat queries.

std::string	activeMigrationTitle
	Identity of the migration currently being rendered. Set by `MigrationManager::ApplySingleMigration` before `ToSql` is called for each step so warnings (e.g. `lup-truncate`) can attribute themselves to the offending migration. Empty / zero when rendering happens outside a migration run (e.g. unit tests calling `ToSql` directly).

std::uint64_t	activeMigrationTimestamp = 0
	Timestamp of the migration currently being rendered (0 outside a migration run).

Detailed Description

Opt-in behavioural knobs that the ToSql migration-plan renderer honours.

When a profile declares a compat: flag (see Lightweight::Config::CompatFlags), dbtool / dbtool-gui construct one of these and pass it alongside the formatter so migration rendering can diverge from strict behaviour per-run.

The context is mutable across successive ToSql calls within the same migration run: it accumulates a column-width cache populated from CreateTable / AlterTable steps as they are rendered, so later Insert / Update steps can truncate oversize values without a server round-trip.

Default-constructed instances are strict (no truncation, no logging); this is the shape used by code paths that don't care about compat, so existing callers keep their behaviour.

Definition at line 573 of file MigrationPlan.hpp.

Member Enumeration Documentation

◆ WidthUnit

enum class Lightweight::MigrationRenderContext::WidthUnit : uint8_t

strong

How a column's declared width is counted. The truncation layer must match the server's interpretation: a VARCHAR(100) on MSSQL holds 100 bytes, but an NVARCHAR(100) holds 100 characters. Truncating to 100 characters when the server budget is 100 bytes still overflows on multi-byte source data.

Definition at line 598 of file MigrationPlan.hpp.

Member Data Documentation

◆ lupTruncate

bool Lightweight::MigrationRenderContext::lupTruncate = false

When true, string values in Insert / Update steps are truncated to the destination column's declared character width and the truncation is logged via SqlLogger::OnWarning. Mirrors LUpd's client-side clipping behaviour; see Lightweight::Config::CompatFlagLupTruncate.

Definition at line 579 of file MigrationPlan.hpp.

◆ columnWidths

std::map<ColumnKey, ColumnWidth> Lightweight::MigrationRenderContext::columnWidths

Cache of declared character widths, populated lazily as plan elements render.

Definition at line 611 of file MigrationPlan.hpp.

◆ widthLookup

std::function<void(MigrationRenderContext&, std::string_view , std::string_view )> Lightweight::MigrationRenderContext::widthLookup

Optional fallback that fetches column widths for a (schema, table) from the live database when the cache has no entry for it.

The renderer calls this once per (schema, table) it can't satisfy from columnWidths. Implementations should issue a single INFORMATION_SCHEMA.COLUMNS (or equivalent) query and write every char/varchar column's declared width back into columnWidths. They may also write a sentinel (e.g. an empty entry) when the table doesn't exist, so the renderer doesn't keep retrying. Default-empty means "no fallback" — i.e. only widths declared by the same migration plan are considered.

Definition at line 623 of file MigrationPlan.hpp.

◆ lookupAttempted

std::set<TableKey> Lightweight::MigrationRenderContext::lookupAttempted

Tables for which widthLookup has already been called; prevents repeat queries.

Definition at line 635 of file MigrationPlan.hpp.

◆ activeMigrationTitle

std::string Lightweight::MigrationRenderContext::activeMigrationTitle

Identity of the migration currently being rendered. Set by MigrationManager::ApplySingleMigration before ToSql is called for each step so warnings (e.g. lup-truncate) can attribute themselves to the offending migration. Empty / zero when rendering happens outside a migration run (e.g. unit tests calling ToSql directly).

Definition at line 642 of file MigrationPlan.hpp.

◆ activeMigrationTimestamp

std::uint64_t Lightweight::MigrationRenderContext::activeMigrationTimestamp = 0

Timestamp of the migration currently being rendered (0 outside a migration run).

Definition at line 644 of file MigrationPlan.hpp.

The documentation for this struct was generated from the following file:

Lightweight/SqlQuery/MigrationPlan.hpp

Classes

Public Types

Public Attributes