Skip to the content
Documentation

Help center

Help for Developers

Build on it: extensions, templates, settings, the APIs, the command line and upgrades.

6101 pages

Start here

Read these pages in this order.

  1. Mit Git entwickeln: vom Benutzer zum Entwickler Der empfohlene Weg, Websites mit Exponential zu bauen. Installieren Sie den Kernel und jede se7enxweb-Erwei...
  2. Erweiterungsentwicklung: das Buch (hier beginnen) Ein vollständiges Buch über den Weg vom Benutzer zum Builder, Erweiterungsentwickler und Theme-Entwickler m...
  3. Benutzerverwaltung von der Konsole Der Anfang von allem rund um die Befehle exp:user in Exponential 6.0.15 - Benutzerkonten auflisten, anlegen...
  4. Das Exponential-Projekt: Hier beginnen Eine Übersicht über das Exponential-Projekt, wie seine Website exponential.earth es darstellt, übertragen a...
  5. Eine Datenbanktabelle anlegen Eine Migration, die eine eigene Tabelle der Site auf jeder Engine aus einer Definition im Format von share/...
  6. Integratoren: hier beginnen Die Einstiegsseite für Systemintegratoren und Integrationsentwickler, die Exponential mit CRM, ERP, PIM, DA...

Good to know

  • Migrationen

    Der Anfang von allem rund um Migrationen in Exponential 6.0.15 - Schema- und Inhaltsänderungen, einmal als Dateien unter Versionskontrolle geschrieben und auf jede Installation einer Site in derselben Reihenfolge…

  • Migrationen: eine erste PHP-Migration, Schritt für Schritt

    In diesem Tutorial schreiben Sie zwei PHP-Migrationen für eine Site: Die eine legt eine Tabelle exp_news_note mit einer ersten Zeile an, die andere veröffentlicht einen Ordner „News“ unter der Inhaltswurzel und vermerkt…

  • Eine Datenbanktabelle anlegen

    Eine Migration, die eine eigene Tabelle der Site auf jeder Engine aus einer Definition im Format von share/db_schema.dba anlegt, eine erste Zeile schreibt und sie in down() wieder löscht - mit Probelauf, Lauf, Prüfung…

  • Migrationen testen: Prüfung, Simulation, Diff und CI

    Vier Ebenen von Tests für Migrationen - die Prüfung ohne Datenbank, der Probelauf, die Simulation mit Diff auf PostgreSQL oder SQLite und ein PHPUnit-Test auf einer Wegwerf-SQLite-Datenbank - und wie man sie in der CI…

  • Mit Migrationen ausliefern: CI/CD und GitHub Actions

    Migrationen bei jeder Auslieferung ausführen - die Reihenfolge der Schritte, ein vollständiger GitHub-Actions-Workflow, der Pull Requests prüft und über SSH ausliefert, die Exit-Codes, auf die eine Pipeline reagiert…

  • Teamarbeit mit Git: Branches, Konflikte und Reihenfolge

    Wie mehrere Entwicklerinnen und Entwickler Migrationen auf Branches schreiben, ohne sich in die Quere zu kommen - Zeitstempel und Reihenfolge, requires für echte Abhängigkeiten, zwei Migrationen mit demselben…

Search by topic

Each button lists the pages of one topic for Developers.

What changed in the source code most recently: the main branch of se7enxweb/exponential, read from its public feed by this server.

Hide merge commits

  1. Updated: The clustering documentation and the file.ini comments name … (opens in a new window) ·
    Message
    Updated: The clustering documentation and the file.ini comments name the kernel's Oracle backend
    
    The pages for database administrators and hosts, the database support overview and the
    PostgreSQL clustering page named the ezoracle extension's backend as the only one for
    Oracle. They now list expDFSFileHandlerOracleBackend with its tables, its gateway dfsoracle
    and the connect string in DBName, keep the extension's backend as the one it replaces with
    the upgrade file, and say that cluster_maintenance works with every kernel backend, in
    English and German. The "More engines" table marks Oracle as built: every database
    Exponential supports has a cluster backend in the kernel. settings/file.ini describes the
    backend's settings.
    a933b21fe9
  2. Updated: exp:cluster:enable makes a one-node cluster on an Oracle sit… (opens in a new window) ·
    Message
    Updated: exp:cluster:enable makes a one-node cluster on an Oracle site, with the cluster tables in the site's own schema
    
    The command told an Oracle installation that it has no cluster backend. With the kernel's
    Oracle backend it now picks it by itself (or with --backend=oracle): the tables of
    kernel/sql/oracle/cluster_dfs_schema.sql in the site's schema through the site's database
    driver, the settings with the site's Database as the connect string (no host or port) and
    its account, and config.cluster.php for the dfsoracle gateway. An Oracle site without an
    account is told its settings are incomplete. The ten schema statements were run as the
    command splits them against Oracle Database Free 26ai. The pages on the cluster commands
    (English and German) list Oracle.
    86146ca46d
  3. Added: The kernel clusters on Oracle: a DFS metadata backend, its tab… (opens in a new window) ·
    Message
    Added: The kernel clusters on Oracle: a DFS metadata backend, its tables, the upgrade of the ezoracle cluster table and its file gateway, with the parity tests every backend passes
    
    A site on several web servers can keep its DFS cluster metadata in Oracle Database 12.1 or
    later with only the oci8 extension. The backend replaces the ezoracle extension's
    eZDFSFileHandlerOracleBackend for clustering (the extension stays what a site needs for
    Oracle as its own database), and one SQL file upgrades the table that backend made. It
    follows the structure of the kernel's PostgreSQL backend, after the ezpostgresqlcluster
    extension by eZ Systems and 7x's fork of it, with the Oracle specifics of the ezoracle
    extension's cluster backend by eZ Systems, and is written method for method against the
    MySQLi and SQLite backends.
    
    - expDFSFileHandlerOracleBackend: oci8 with every value bound (no bind is named after a
      column: SIZE is reserved, ORA-01745); a new connection per backend or a persistent one
      (DBPersistentConnection); AL32UTF8. An Oracle UPDATE waits for a row lock without limit,
      so every change locks its rows first with SELECT ... FOR UPDATE WAIT DBLockTimeout; a
      wait that ran out (ORA-30006, ORA-00054 from 23ai on), a deadlock (ORA-00060), a
      serialization failure (ORA-08177) and the duplicate key of two concurrent upserts
      (ORA-00001) are retried up to DBExecuteRetries, a transaction written whole or not at
      all. MERGE upserts; a plain INSERT whose duplicate key means "being generated" for the
      one-winner cache generation; take-overs conditioned on the mtime read; purges with FOR
      UPDATE SKIP LOCKED; case sensitive LIKE ... ESCAPE '\' name searches that use the name
      index, {a,b} wildcards expanded; OFFSET ... FETCH NEXT for the lists. The size column is
      filesize and is returned as size; Oracle's NULL for the empty string is read as '', and
      a list of every scope but some keeps a row without scope. Cluster events.
    - kernel/sql/oracle/cluster_dfs_schema.sql…
    c4c86090b5
  4. Updated: The clustering documentation and the file.ini comments name … (opens in a new window) ·
    Message
    Updated: The clustering documentation and the file.ini comments name the kernel's MongoDB backend
    
    The pages for database administrators and hosts, the database support overview and the
    PostgreSQL clustering page said MongoDB had no cluster backend yet. They now list
    expDFSFileHandlerMongoDBBackend with its collections file, its gateway dfsmongodb, the
    replica set connection string in DBHost and a link to the feature page, in English and
    German; the "More engines" table marks MongoDB as built and leaves Oracle as the one that
    follows. settings/file.ini describes the backend's settings next to the other engines'.
    398b9e5935
  5. Updated: exp:cluster:enable makes a one-node cluster on a MongoDB sit… (opens in a new window) ·
    Message
    Updated: exp:cluster:enable makes a one-node cluster on a MongoDB site, with the cluster collections in the site's own database
    
    The command told a MongoDB installation that it has no cluster backend. With the kernel's
    MongoDB backend it now picks it by itself (or with --backend=mongodb): the collections and
    indexes of kernel/sql/mongodb/cluster_dfs_schema.json in the site's database, created by the
    backend, the settings with the site's server, account and database, and config.cluster.php
    for the dfsmongodb gateway. A server without authentication (no User) is accepted. The
    backend list now names each engine's schema file, so an engine whose schema is not SQL is set
    up by its backend; the second half of the switch (settings, moving the files in, the marker)
    is shared by both paths. Checked against a throwaway MongoDB 8.0: collections created and
    counted, documents of another origin refused, those of an earlier switch of ours emptied. The
    pages on the cluster commands (English and German) list MongoDB.
    ae9d56211c
  6. Added: The kernel clusters on MongoDB: a DFS metadata backend, its co… (opens in a new window) ·
    Message
    Added: The kernel clusters on MongoDB: a DFS metadata backend, its collections and its file gateway, with the parity tests every backend passes
    
    A site on MongoDB can keep its DFS cluster metadata in its own database, on one server or a
    replica set, without an extension and without the PHP library: the backend only needs the
    mongodb extension's driver classes. It follows the structure of the kernel's PostgreSQL
    backend, after the ezpostgresqlcluster extension by eZ Systems and 7x's fork of it, and is
    written method for method against the SQLite and MySQLi backends.
    
    - expDFSFileHandlerMongoDBBackend: one document per file with _id = name_hash; the unique
      _id is the lock of a cache generation (one insert wins, the duplicate key means "being
      generated"); take-overs and timeout checks are one update conditioned on the mtime read;
      deletes mark files expired with an update pipeline (mtime -ABS(mtime)); upserts for
      stores. No multi-document transactions, so a single mongod works as well as a replica
      set; writes of several documents are ordered so that a stopped process leaves a state the
      next one recovers from (the file before its document, the new name before the old one is
      removed). Network errors, failovers and writes the server labels retryable are tried
      again up to DBExecuteRetries; finding a server waits DBLockTimeout at most. LIKE patterns
      and shell wildcards become anchored, case sensitive regular expressions that use the name
      index, with every regular expression character in a name a literal. The account is passed
      as driver options, never inside a connection string; DBHost may hold a whole mongodb://
      string for a replica set. Cluster events as on the other engines.
    - kernel/sql/mongodb/cluster_dfs_schema.json: the collections ezdfsfile and ezdfsfile_cache
      and their indexes, created by expDFSFileHandlerMongoDBBackend::createSchema() (existing
      ones are left alone); 64 bit sizes and times.
    - The gateway kernel/clustering/dfsmongodb.php f…
    b12d07438d
  7. Updated: The clustering documentation and the file.ini comments name … (opens in a new window) ·
    Message
    Updated: The clustering documentation and the file.ini comments name the kernel's SQLite backend, for one machine
    
    - settings/file.ini lists expDFSFileHandlerSQLiteBackend with its DBName rule, its schema
      file and DBLockTimeout, and says that the server settings are not used.
    - The clustering pages for database administrators and hosting (English and German) have a
      row for the SQLite backend: one machine only, for a one-node cluster and tests.
    - The database support page names the SQLite cluster backend in its table and in the
      SQLite section.
    - The PostgreSQL cluster pages mark SQLite as built among the further engines.
    001e4bfbd9
  8. Updated: exp:cluster:enable makes a one-node cluster on a SQLite site… (opens in a new window) ·
    Message
    Updated: exp:cluster:enable makes a one-node cluster on a SQLite site, with the cluster tables in the site's own database file
    
    The command told a SQLite installation that it has no cluster backend. With the kernel's
    SQLite backend it now picks it by itself (or with --backend=sqlite): the tables of
    kernel/sql/sqlite/cluster_dfs_schema.sql in the site's database file, the settings with the
    file as DBName and no host, port or account, and config.cluster.php for the dfssqlite
    gateway. --backend accepts every engine the kernel has a backend for, and the message for an
    engine without one lists them. Checked: the schema statements run through the site's own
    SQLite driver on a throwaway database, and the backend then reads and writes that file. The
    pages on the cluster commands (English and German) list SQLite, one node only.
    735942af77
  9. Added: The kernel clusters one node on SQLite: a DFS metadata backend… (opens in a new window) ·
    Message
    Added: The kernel clusters one node on SQLite: a DFS metadata backend, its tables and its file gateway, with the parity tests every backend passes
    
    A site whose PHP servers share one host can keep its DFS cluster metadata in a SQLite file
    without an extension or a database server, and the cluster tests run with nothing to
    install. SQLite is one file on one machine, so the backend is for a one-node cluster and for
    tests, never for several web servers. It follows the structure of the kernel's PostgreSQL
    backend, after the ezpostgresqlcluster extension by eZ Systems and 7x's fork of it.
    
    - expDFSFileHandlerSQLiteBackend, method for method against the MySQLi backend: PDO with
      bound parameters; every transaction BEGIN IMMEDIATE in place of row locks; the wait for
      the write lock bounded by the busy timeout (DBLockTimeout) and SQLITE_BUSY/SQLITE_LOCKED
      retried up to DBExecuteRetries, a transaction's work never run outside it; WAL journal
      mode; ON CONFLICT upserts; case sensitive GLOB name searches that use the name index,
      with {a,b} wildcards expanded; the database file opened without the create flag; cluster
      events. PHP 8.4+ uses the Pdo\Sqlite constants, so PHP 8.5 reports no deprecation.
    - kernel/sql/sqlite/cluster_dfs_schema.sql (64 bit sizes and times) and the gateway
      kernel/clustering/dfssqlite.php for CLUSTER_STORAGE_BACKEND dfssqlite.
    - Tests: a DB-less unit test, and the parity checks on a SQLite file, 29 of 29, among them
      four processes generating one cache key and a lock held by another connection. Neither
      needs a server; the CI job of the cluster backends runs both. The parity checks let an
      engine without row locks say how a second connection holds the lock (PostgreSQL 18 27/27
      and MySQL 8.0 23 with 2 skipped by design, as before).
    - Verified on alpha through the real eZDFSFileHandler: one node (2 MB file, copy, rename,
      delete, purge, view cache and its name trunk expiry), index_cluster.php with dfssqlite
      (whole file, byte range, 404)…
    7d331c5f14
  10. Added: The feature page on queries over id sets of any size in Englis… (opens in a new window) ·
    Message
    Added: The feature page on queries over id sets of any size in English and German, and its 6.0.15 changelog entries
    
    The page describes expDBInList and site.ini [DatabaseSettings] InListChunkSize, the limits of
    each engine, the kernel code that uses the helper and the places that were checked and left with
    the reason. It credits Ekkehard Dörre, whose pull request #365 fixed the role page on Oracle and
    brought the pattern and its test. Both feature indexes link it, the German path map maps it, and
    the 6.0.15 changelog has a section of its own.
    7cb1288e75

Showing 1–10 of 203 commits

Recently updated

Free documentation: use it to help yourself and others

This help is free documentation, licensed under the GNU Free Documentation License 1.3 or later. You may use it, share it and improve it:

  • Study it Read any page as often as you like, for any purpose.
  • Copy it and share it Give pages to other users, on paper or as files. Keep the copyright notice, the license notice and a copy of the license with every copy.
  • Translate, adapt and improve it Share your version under the same license, with the original notices and a note of what you changed.
  • Cite it Refer to a page by its title and its address in the help, or its file in the doc directory.
  • Contribute fixes back Use "Edit on GitHub" on any page. Changes are accepted under the same license, so they stay free for everyone.

Exponential Copyright © 1998-2026 Exponential Foundation and 7x