gudusoft.gsqlparser.dlineage.util

Class NormalNameCollisions

Object

gudusoft.gsqlparser.dlineage.util.NormalNameCollisions

public final class NormalNameCollisions
extends Object

Keeps two different names from sharing one lineage normal name.

Lineage keys its models by SQLUtil#normalizeIdentifier, and the legacy fold behind it can give two different names one string: PostgreSQL "Amount" and amount both become AMOUNT, ClickHouse userId and userid both become USERID. Lineage then treats the two as one column, and every reference to either picks up both as its source. The same string is also displayed in places, so it cannot change for every name.

So before the first statement is analyzed, scan(gudusoft.gsqlparser.EDbVendor, java.util.Collection<java.lang.String>) reads every identifier token of the analysis's SQL and finds the legacy strings that more than one identity shares. Only a name whose legacy string is contested changes (normalName(java.util.Set<java.lang.String>, gudusoft.gsqlparser.EDbVendor, gudusoft.gsqlparser.sqlenv.ESQLDataObjectType, java.lang.String, java.lang.String)): it keeps the legacy string when that string, read back as SQL, denotes it (amount keeps AMOUNT on PostgreSQL), and otherwise gets a spelling that denotes it ("Amount"). Lineage compares these strings as SQL names, so each one must name its own identity; the first name to arrive keeping the legacy string would not, because AMOUNT read back as SQL is amount whoever holds it. Column names are contested without case, because lineage matches column normal names case-insensitively even where the legacy fold keeps case (Oracle, whose quoted "Amount" gives Amount next to a bare amount's AMOUNT). Names are told apart under the analysis's own rules, so a configured profile's stricter ones count.

The result is a pure function of the contested set, so all spellings of a legacy string change together. A name the scan did not see (dynamic SQL, a name only metadata holds) follows the same rule: it changes if its string is contested, and otherwise keeps the merge the legacy fold always made. No string is ever split for some of its names and merged for the rest.

Method Summary

Modifier and Type	Method and Description
static String	contestedKey(IdentifierService service, ESQLDataObjectType type, String legacy) The contested key of the legacy name legacy of a name of role type, as scan(gudusoft.gsqlparser.EDbVendor, java.util.Collection<java.lang.String>) returns them: for a caller that finds two names sharing one legacy name outside the SQL text a scan reads.
static Set<String>	legacyNames(Set<String> contested) The legacy strings of contested, whatever their class and upper-cased: a cheap first test, which normalName(java.util.Set<java.lang.String>, gudusoft.gsqlparser.EDbVendor, gudusoft.gsqlparser.sqlenv.ESQLDataObjectType, java.lang.String, java.lang.String) then makes exact.
static String	normalName(Set<String> contested, EDbVendor vendor, ESQLDataObjectType type, String origin, String legacy) The normal name of one identifier segment, given the contested keys of its analysis.
static String	normalName(Set<String> contested, IdentifierService service, ESQLDataObjectType type, String origin, String legacy) normalName(Set, EDbVendor, ESQLDataObjectType, String, String) under the analysis's own identifier rules; pass the service scan(IdentifierService, Collection) used.
static Set<String>	scan(EDbVendor vendor, Collection<String> sqlTexts) The legacy normal names that two different identities share in sqlTexts, keyed for normalName(java.util.Set<java.lang.String>, gudusoft.gsqlparser.EDbVendor, gudusoft.gsqlparser.sqlenv.ESQLDataObjectType, java.lang.String, java.lang.String).
static Set<String>	scan(IdentifierService service, Collection<String> sqlTexts) scan(EDbVendor, Collection) under the analysis's own identifier rules, which a configured profile may make stricter than the vendor's (SQL Server under a case-sensitive collation tells Amount from AMOUNT).
static Set<String>	scanTokens(IdentifierService service, Collection<TSourceTokenList> tokenLists) scan(gudusoft.gsqlparser.EDbVendor, java.util.Collection<java.lang.String>) over tokens already read, such as a parser's token list after it parsed the analysis's only unit, which spares lexing the text a second time.

Methods inherited from class Object

clone, equals, finalize, getClass, hashCode, notify, notifyAll, toString, wait, wait, wait

Method Detail

scan

public static Set<String> scan(EDbVendor vendor,
                               Collection<String> sqlTexts)

The legacy normal names that two different identities share in sqlTexts, keyed for normalName(java.util.Set<java.lang.String>, gudusoft.gsqlparser.EDbVendor, gudusoft.gsqlparser.sqlenv.ESQLDataObjectType, java.lang.String, java.lang.String). Empty when no two names collide, which is the common case.

Parameters:

vendor - the analysis's vendor; names are compared under its default identifier rules (see scan(IdentifierService, Collection) for a configured profile)

sqlTexts - every unit of SQL the analysis will read

Returns:

the contested keys

scan

public static Set<String> scan(IdentifierService service,
                               Collection<String> sqlTexts)

scan(EDbVendor, Collection) under the analysis's own identifier rules, which a configured profile may make stricter than the vendor's (SQL Server under a case-sensitive collation tells Amount from AMOUNT).

scanTokens

public static Set<String> scanTokens(IdentifierService service,
                                     Collection<TSourceTokenList> tokenLists)

scan(gudusoft.gsqlparser.EDbVendor, java.util.Collection<java.lang.String>) over tokens already read, such as a parser's token list after it parsed the analysis's only unit, which spares lexing the text a second time.

legacyNames

public static Set<String> legacyNames(Set<String> contested)

The legacy strings of contested, whatever their class and upper-cased: a cheap first test, which normalName(java.util.Set<java.lang.String>, gudusoft.gsqlparser.EDbVendor, gudusoft.gsqlparser.sqlenv.ESQLDataObjectType, java.lang.String, java.lang.String) then makes exact.

normalName

public static String normalName(Set<String> contested,
                                EDbVendor vendor,
                                ESQLDataObjectType type,
                                String origin,
                                String legacy)

The normal name of one identifier segment, given the contested keys of its analysis.

Parameters:

contested - keys from scan(gudusoft.gsqlparser.EDbVendor, java.util.Collection<java.lang.String>)

vendor - vendor the legacy name was folded for

type - role of the segment

origin - the segment as written

legacy - its legacy normal name

Returns:

legacy, unless another name shares it and it does not denote origin

normalName

public static String normalName(Set<String> contested,
                                IdentifierService service,
                                ESQLDataObjectType type,
                                String origin,
                                String legacy)

normalName(Set, EDbVendor, ESQLDataObjectType, String, String) under the analysis's own identifier rules; pass the service scan(IdentifierService, Collection) used.

contestedKey

public static String contestedKey(IdentifierService service,
                                  ESQLDataObjectType type,
                                  String legacy)

The contested key of the legacy name legacy of a name of role type, as scan(gudusoft.gsqlparser.EDbVendor, java.util.Collection<java.lang.String>) returns them: for a caller that finds two names sharing one legacy name outside the SQL text a scan reads.