unogenerator-1.4.0. Mejorada la estabilidad de la concurrencia

GEMINI.md

@@ -16,7 +16,15 @@
 **Decision:**
 - **Safety-First Optimization:** The optimization now only skips style application if both the intended style is `Default` AND the document's default is also `Default`. If a custom `default_cell_style` (like `Normal`) is defined, it will always be explicitly applied to ensure document consistency.
 
-### 3. Regression Testing
+### 5. Localization Workflow (`poe translate`)
+**Observation:** The project uses `poe translate` (linked to `unogenerator.poethepoet:translate`) to automate the synchronization of `.po` files with the `.pot` template.
+**Experience:** 
+- **English Redundancy:** `en.po` has been removed as it was redundant. Since the source strings in the code are in English, the system now uses `fallback=True` in `gettext.translation` calls within `demo.py`. This ensures that when the English locale is requested, it falls back to the original English strings without needing a separate `.po` file.
+- **Header Fragility:** Running `poe translate` may overwrite manual header improvements or reset them to defaults.
+- **Multilingual Support:** While efficient for syncing Spanish (the developer's primary language), it might leave other languages (French, Romanian) with empty `msgstr` entries if not carefully monitored.
+- **Verification:** After running `poe translate`, always verify that all message strings in all supported languages are correctly populated and that headers maintain the correct project metadata and dates. Manual restoration of translations for non-primary languages may be required if the tool only targets one language.
+
+### 6. Regression Testing
 New tests have been added to `tests/test_unogenerator.py` to protect these fixes:
 - `test_ods_row_height_consistency`: Ensures heights stay at 452 even after `setColumnsWidth` with many columns.
 - `test_ods_normal_style_applied`: Verifies that `Normal` style is correctly applied by `ODS_Standard`.
@@ -27,3 +35,21 @@ New tests have been added to `tests/test_unogenerator.py` to protect these fixes
 **Decision:**
 - **Explicit Dependency:** `envwrap` has been added to `pyproject.toml`. While not a direct dependency of the library's core logic, it is essential for the environment's stability when `tqdm` and `uno` coexist.
 - **Safety:** It is a safe, lightweight utility for environment variable wrapping. Adding it explicitly prevents the intermittent `ImportError` and ensures that tests and demo scripts run reliably across different setups.
+
+### 7. PyUNO Initialization in Concurrent Processes
+**Problem:** When using `multiprocessing` with the `spawn` start method (as used in the demo), child processes would fail with `SystemError: pyuno runtime is not initialized`. This is because `uno.getComponentContext()` (from the standard `uno.py`) returns a cached context object that is initialized during the module's import phase. In `spawn` mode, this initialization happens during the child process's bootstrap and is not valid for the subsequent execution phase.
+
+**Decision:**
+- **Direct PyUNO Access:** We now use `pyuno.getComponentContext()` directly instead of the cached `uno.getComponentContext()`.
+- **Deferred Initialization:** In `unogenerator.py`, we defined a local `getComponentContext()` wrapper that calls the underlying `pyuno` function. This ensures that every time the context is requested (e.g., during `ODF.__init__` or `createUnoService`), `pyuno` performs a fresh initialization check in the current process phase.
+- **Verification:** A new test `tests/test_concurrency.py` has been added to specifically verify that `spawn`-ed processes can successfully initialize and use UNO objects.
+
+### 8. Optimized Thread-Safety via Instance-Based Serialization
+**Problem:** Running the demo with concurrent threads and a shared LibreOffice server (`COMMONSERVER_CONCURRENT_THREADS`) was unstable due to non-thread-safe sharing of a single connection. However, using a single global lock serialized execution even for workers using independent LibreOffice instances, causing unnecessary performance loss.
+
+**Decision:**
+- **Instance-Based Reentrant Lock:** Each `LibreofficeServer` object now maintains its own `threading.RLock()`.
+- **Smart Synchronization:** The `@uno_safe` class decorator dynamically retrieves the lock from the document's associated server instance (`self.server._lock`).
+- **Parallel Performance:** This allows workers using independent servers (different ports) to run at full parallel speed without waiting for each other.
+- **Shared-Server Stability:** Workers sharing the same `LibreofficeServer` instance will correctly share the same lock, ensuring serialized access to the single connection and maintaining stability.
+- **Internal Protection:** Global UNO bridge calls (like `getComponentContext`) remain protected by a global `_uno_bridge_lock` to ensure the process-wide PyUNO state is not corrupted during initialization.

pyproject.toml

@@ -1,6 +1,6 @@
 [project]
 name = "unogenerator"
-version = "1.3.0"
+version = "1.4.0"
 description = "Libreoffice files generator programmatically with python and Libreoffice server instances"
 authors = [
     { name = "turulomio", email = "turulomio@yahoo.es" }

unogenerator/__init__.py

@@ -9,8 +9,8 @@
 logger.addHandler(logging.NullHandler()) # Add NullHandler to prevent "No handlers could be found" message
 logger.setLevel(logging.WARNING) # Set default level for the library to WARNING
 
-__version__ = '1.3.0'
-__versiondatetime__=datetime(2026, 5, 24, 9, 52)
+__version__ = '1.4.0'
+__versiondatetime__=datetime(2026, 6, 7, 4, 20)
 __versiondate__=__versiondatetime__.date()
 
 

unogenerator/demo.py

@@ -1,14 +1,11 @@
-from uno import getComponentContext
-
-getComponentContext()
 import argparse
 from collections import OrderedDict
 from concurrent.futures import ProcessPoolExecutor, as_completed,  ThreadPoolExecutor
 from datetime import datetime, date, timedelta
 from gettext import translation # Removed 'info'
 import logging # Import logging module
 from importlib.resources import files
-from os import system
+from subprocess import run
 from pydicts.currency import Currency
 from pydicts.percentage import Percentage
 from unogenerator import ODT_Standard, ODS_Standard, __version__,  commons, ColorsNamed, Coord, LibreofficeServer, helpers, types, Range
@@ -20,8 +17,6 @@
 except:
     _=str
 
-type_choices=[ "SEQUENTIAL",  "CONCURRENT_PROCESS",  "CONCURRENT_THREADS", "COMMONSERVER_SEQUENTIAL","COMMONSERVER_CONCURRENT_PROCESS","COMMONSERVER_CONCURRENT_THREADS" ]
-
 ## If arguments is None, launches with sys.argc parameters. Entry point is toomanyfiles:main
 
 logger = logging.getLogger(__name__) # Get logger for this module
@@ -68,28 +63,29 @@
 
 
 
-## You can call with main(['--pretend']). It's equivalento to os.system('program --pretend')
+## You can call with main(['--pretend']). It's equivalento to run(['program', '--pretend'])
 ## @param arguments is an array with parser arguments. For example: ['--argument','9']. 
 def demo(arguments=None):
+
     parser=argparse.ArgumentParser(prog='unogenerator', description=_('Create example files using unogenerator module'), epilog=commons.argparse_epilog(), formatter_class=argparse.RawTextHelpFormatter)
     parser.add_argument('--version', action='version', version=__version__)
     parser.add_argument('--debug', help=_("Debug program information"), choices=["DEBUG","INFO","WARNING","ERROR","CRITICAL"], default="ERROR")
     group= parser.add_mutually_exclusive_group(required=True)
     group.add_argument('--create', help="Create demo files", action="store_true",default=False)
     group.add_argument('--remove', help="Remove demo files", action="store_true", default=False)
     group.add_argument('--benchmark', help="Executes all types to compare its benchmark", action="store_true", default=False)
-    parser.add_argument('--type', help="Debug program information", choices=type_choices,  default="COMMONSERVER_CONCURRENT_PROCESS")
+    parser.add_argument('--type', help="Debug program information", choices=[t.name for t in types.DemoType],  default=types.DemoType.CONCURRENT_PROCESS.name)
     args=parser.parse_args(arguments)
     commons.addDebugSystem(args.debug)
-    demo_command(args.create, args.remove, args.benchmark, args.type)
+    demo_command(args.create, args.remove, args.benchmark, types.DemoType[args.type])
 
 def demo_command(create, remove, benchmark, type):
     languages=['es', 'en',  'ro',  'fr']
 
     if benchmark is True:
-        for type in type_choices:
+        for t in types.DemoType:
             #demo_command(True, False,  False,  type)
-            system(f"unogenerator_demo --create --type {type}")
+            run(["unogenerator_demo", "--create", "--type", t.name], check=True)
 
     if remove==True:
             for language in languages:
@@ -102,10 +98,10 @@ def demo_command(create, remove, benchmark, type):
 
     if create==True:
         start=datetime.now()
-        instances=3
+        instances=4
         total_documents=len(languages)*2
 
-        if type=="CONCURRENT_PROCESS":            
+        if type==types.DemoType.CONCURRENT_PROCESS:            
             futures=[]
             print(_("Launching demo with {0} workers without common server using concurrent processes").format(instances))
 
@@ -127,7 +123,7 @@ def demo_command(create, remove, benchmark, type):
                 result = future.result()
                 results.append(result)    
 
-        elif type=="COMMONSERVER_CONCURRENT_PROCESS":            
+        elif type==types.DemoType.COMMONSERVER_CONCURRENT_PROCESS:            
             futures=[]
             print(_("Launching demo with {0} workers with common server using concurrent processes").format(instances))
 
@@ -160,7 +156,7 @@ def demo_command(create, remove, benchmark, type):
             finally:
                 main_server.stop() # Ensure the main server is stopped when done
 
-        elif type=="CONCURRENT_THREADS":            
+        elif type==types.DemoType.CONCURRENT_THREADS:            
             futures=[]
             print(_("Launching demo with {0} workers without common server using concurrent threads").format(instances))
 
@@ -182,7 +178,7 @@ def demo_command(create, remove, benchmark, type):
                     result = future.result()
                     results.append(result)
 
-        elif type=="COMMONSERVER_CONCURRENT_THREADS":            
+        elif type==types.DemoType.COMMONSERVER_CONCURRENT_THREADS:            
             futures=[]
             print(_("Launching demo with {0} workers with common server using concurrent threads").format(instances))
 
@@ -205,7 +201,7 @@ def demo_command(create, remove, benchmark, type):
                     result = future.result()
                     results.append(result)
 
-        elif type=="COMMONSERVER_SEQUENTIAL":
+        elif type==types.DemoType.COMMONSERVER_SEQUENTIAL:
             with LibreofficeServer() as server:
                 print(_("Launching demo with one common server sequentially"))
                 with tqdm(total=total_documents) as progress:
@@ -215,7 +211,7 @@ def demo_command(create, remove, benchmark, type):
                         demo_odt_standard(language,  server)       
                         progress.update()
 
-        elif type=="SEQUENTIAL":
+        elif type==types.DemoType.SEQUENTIAL:
                 print(_("Launching demo without one common server sequentially"))
                 with tqdm(total=total_documents) as progress:
                     for language in languages:
@@ -228,7 +224,7 @@ def demo_command(create, remove, benchmark, type):
 
 
 def demo_ods_standard(language, server):
-    lang1=translation('unogenerator', files("unogenerator") / 'locale', languages=[language])
+    lang1=translation('unogenerator', files("unogenerator") / 'locale', languages=[language], fallback=True)
     lang1.install()
     _=lang1.gettext
 
@@ -264,7 +260,7 @@ def demo_ods_standard(language, server):
 
 
 def demo_odt_standard(language, server):
-    lang1=translation('unogenerator', files("unogenerator") / 'locale', languages=[language])
+    lang1=translation('unogenerator', files("unogenerator") / 'locale', languages=[language], fallback=True)
     lang1.install()
     _=lang1.gettext
 

unogenerator/locale/es.po

@@ -1,23 +1,21 @@
 # Spanish translations for unogenerator package
-# Traducciones al español para el paquete unogenerator.
-# Copyright (C) 2024 turulomio
+# Copyright (C) 2020-2026 Mariano Muñoz
 # This file is distributed under the same license as the unogenerator package.
+# Mariano Muñoz <turulomio@yahoo.es>, 2015-2026.
 #
-# Mariano Muñoz <turulomio@yahoo.es>, 2015, 2020, 2024.
 msgid ""
 msgstr ""
 "Project-Id-Version: Unogenerator\n"
 "Report-Msgid-Bugs-To: \n"
-"POT-Creation-Date: 2026-05-24 09:56+0200\n"
-"PO-Revision-Date: 2024-07-31 12:00:00+0200\n"
-"Last-Translator: turulomio@yahoo.es\n"
-"Language-Team: turulomio@yahoo.es\n"
+"POT-Creation-Date: 2026-06-07 04:20+0200\n"
+"PO-Revision-Date: 2026-06-07 10:15+0200\n"
+"Last-Translator: Mariano Muñoz <turulomio@yahoo.es>\n"
+"Language-Team: Spanish <turulomio@yahoo.es>\n"
 "Language: es\n"
 "MIME-Version: 1.0\n"
 "Content-Type: text/plain; charset=UTF-8\n"
 "Content-Transfer-Encoding: 8bit\n"
 "Plural-Forms: nplurals=2; plural=(n != 1);\n"
-"X-Generator: Lokalize 19.12.3\n"
 
 msgid "There was a problem converting com.sun.star.util.DateTime to datetime."
 msgstr "Hubo un problema al convertir com.sun.star.util.DateTime a datetime"
@@ -335,8 +333,11 @@ msgstr "Esto es un conjunto de símbolos: .,:;?ºª-/()."
 msgid "unogenerator_documentation_{0}.ods took {1} in {2}"
 msgstr "unogenerator_documentation_{0}.ods tardó {1} en {2}"
 
-msgid "Style name"
-msgstr "Nombre de estilo"
+msgid "Color name"
+msgstr "Nombre del color"
+
+msgid "Hex"
+msgstr "Hex"
 
 msgid "Date and time"
 msgstr "Fecha y hora"
@@ -507,3 +508,6 @@ msgstr "La extensión del nombre del fichero debe ser 'ods'."
 
 msgid "Filename extension must be 'xlsx'."
 msgstr "La extensión del nombre del fichero debe ser 'xlsx'."
+
+#~ msgid "Style name"
+#~ msgstr "Nombre de estilo"