Meerschaum 1¶

Historical release notes for the 1.x series.

1.7.x Releases¶

Read release notes

The 1.7 series was short and sweet with a big focus on improving the web API. The highlight feature was the integrated webterm, and the series includes many bugfixes and improvements.

v1.7.3 – v1.7.4¶

Fix an issue with the local stack healthcheck. Due to some edge cases, the local stack docker-compose.yaml file would not be correctly formatted until edit config had been executed. This patch ensures the files are synced with each invocation of stack.
Fix an issue when running the local stack with non-default ports. Initializing a local stack with a different database port (e.g. 5433) now routes correctly within the Docker compose network (now patching to internal port to 5432).
Fix upgrade mrsm behavior. Recent changes to stack broke the automatic stack pull within mrsm upgrade mrsm.

v1.7.2¶

Fix role "root" does not exist from stack logs. Although the healthcheck was working as expected, the log output was filled with Error FATAL: role "root" does not exist. These errors have been fixed.
Fix MRSM_CONFIG behavior when running start api --production. Starting the Web API through gunicorn (i.e. --production) now respects MRSM_CONFIG. This is useful for running stack up with non-default credentials.
Added --insecure as an alias for --no-auth. To compliment the newly added --secure flag, starting the Web API with --insecure will bypass authentication.
Bump default TimescaleDB version to PG15. The default TimescaleDB version for the Meerschaum stack is now latest-pg15-oss.
Pass sysargs to docker compose via stack This patch allows for jumping into the api container:

mrsm stack exec api bash

Added the API endpoint /healthcheck. This is used to determine reachability and the health of the local stack.

v1.7.0 – v1.7.1¶

Remove get_backtrack_data() for instance connectors. If provided, this method will still override the new generic implementation.
Add --keyfile and --certfile support. When starting the Web API, you may now run via HTTPS with --keyfile and --certfile. Older releases required the keys to be set in MRSM_CONFIG. This also brings SSL support for --production (Gunicorn).
Add the Webterm to the Web Console. At long last, the webterm is embedded within the web console and is accessible from the Web API at the endpoint /webterm. You must provide your active, authorized session ID to access to the Webterm.
Add --secure to start api. Starting the Web API with --secure will now disallow actions from non-administrators. This is recommend for shared deployments.
Fixed the registration page on the Web API. Users should now be able to create accounts from Dockerized deployments.
Held back dash-extensions The recent 1.0.2+ releases have shipped some broken changes, so dash-extensions is held back to 1.0.1 until newer releases have been tested.
Allow for digits in environment connectors. Connectors defined as environment variables may now have digits in the type.

export MRSM_A2B_TEST='{"foo": "bar"}'

Fixed stack on Windows.
Fixed a false error with background jobs.
Increased the minimum password length to 5.

1.6.x Releases¶

Read release notes

The biggest features of the 1.6.x series were all about chunking and adding support for syncing generators. The series was also full of minor bugfixes, contributing to an even more polished experience. It also was the first release to drop support for a Python version, formally deprecating Python 3.7.

v1.6.16 – v1.6.19¶

Add Pydantic v2 support The only feature which requires Pydantic v1 is the --schedule flag, which will throw a warning with a hint to install an older version. The underlying libraries for this feature should have Pydantic v2 support merged soon.
Bump dependencies. This patch bumps the minimum required versions for typing-extensions, rich, prompt-toolkit, rocketry, uvicorn, websockets, and fastapi and loosens the minimum version of pydantic.
Fix shell formatting on Windows 10. Some edge case issues have been patched for older versions of Windows.

v1.6.15¶

Sync chunks in the copy pipes action. This will help with large out-of-memory pipes.

v1.6.14¶

Added healthchecks to mrsm stack up. The internal Docker Compose file for mrsm stack was bumped to version 3.9, and secrets were replaced with environment variable references.
Fixed --no-auth when starting the API. The command mrsm start api --no-auth now correctly handles sessions.

v1.6.13¶

Remove \\u0000 from strings when inserting into PostgreSQL. Replace both \0 and \\u0000 with empty strings when streaming rows into PostgreSQL.

v1.6.12¶

Allow nested chunk generators. This patch more gracefully handles labels for situations with nested chunk generators and adds and explicit test for this scenario.

import meerschaum as mrsm
pipe = mrsm.Pipe('foo', 'bar', instance='sql:memory')

docs = [{'color': 'red'}, {'color': 'green'}]
num_chunks = 3
num_batches = 2

def build_chunks():
    return (
        [
            {'chunk_ix': chunk_ix, **doc}
            for doc in docs
        ]
        for chunk_ix in range(num_chunks)
    )

batches = (
    (
        [
            {'batch_ix': batch_ix, **doc}
            for doc in chunk
        ]
        for chunk in build_chunks()
    )
    for batch_ix in range(num_batches)
)

pipe.sync(batches)
print(pipe.get_data())
#     batch_ix  chunk_ix  color
# 0          0         0    red
# 1          0         0  green
# 2          0         1    red
# 3          0         1  green
# 4          0         2    red
# 5          0         2  green
# 6          1         0    red
# 7          1         0  green
# 8          1         1    red
# 9          1         1  green
# 10         1         2    red
# 11         1         2  green

v1.6.11¶

Fix an issue with in-place syncing. When syncing a SQL pipe in-place with a backtrack interval, the interval is applied to the existing data stage to avoid inserting duplicate rows.

v1.6.9 – v1.6.10¶

Improve thread safety checks. Added checks for IS_THREAD_SAFE to connectors to determine whether to use mutlithreading.
Fix an issue with custom flags while syncing. This patch includes better handling of custom flags added from plugins during the syncing process.

v1.6.8¶

Added as_iterator to Pipe.get_data(). Passing as_iterator=True (or as_chunks) to Pipe.get_data() returns a generator which returns chunks of Pandas DataFrames.

Each DataFrame is the result of a Pipe.get_data() call with intermediate datetime bounds between begin and end of size chunk_interval (default datetime.timedelta(days=1) for time-series / 100,000 IDs for integers).

import meerschaum as mrsm

pipe = mrsm.Pipe(
    'a', 'b',
    columns={'datetime': 'id'},
    dtypes={'id': 'Int64'},
)
pipe.sync([
    {'id': 0, 'color': 'red'},
    {'id': 1, 'color': 'blue'},
    {'id': 2, 'color': 'green'},
    {'id': 3, 'color': 'yellow'},
])

### NOTE: due to non-inclusive end bounding,
###       each chunk contains
###       (chunk_interval - 1) rows.
chunks = pipe.get_data(
    chunk_interval = 2,
    as_iterator = True,
)
for chunk in chunks:
    print(chunk)

#    id color
# 0   0   red
# 1   1  blue
#    id   color
# 0   2   green
# 1   3  yellow

Add server-side cursor support to SQLConnector.read(). If chunk_hook is provided, keep an open cursor and stream the chunks one-at-a-time. This allows for processing very large out-of-memory data sets.

To return the results of the chunk_hook callable rather than a dataframe, pass as_hook_result=True to receive a list of values.

If as_iterator is provided or chunksize is None, then SQLConnector.read() reverts to the default client-side cursor implementation (which loads the entire result set into memory).

import meerschaum as mrsm
conn = mrsm.get_connector()

def process_chunk(df: 'pd.DataFrame', **kw) -> int:
    return len(df)

results = conn.read(
    "very_large_table",
    chunk_hook = process_chunk,
    as_hook_results = True,
    chunksize = 100,
)

results[:2]
# [100, 100]

Remove --sync-chunks and set its behavior as default. Due to the above changes to SQLConnector.read(), sync_chunks now defaults to True in Pipe.sync(). You may disable this behavior with --chunksize 0.

v1.6.7¶

Improve memory usage when syncing generators. To more lazily sync chunks from generators, pool.map() has been replaced with pool.imap().

v1.6.6¶

Issue one ALTER TABLE query per column for SQLite, MSSQL, DuckDB, and Oracle SQL. SQLite and other flavors do not support multiple columns in an ALTER TABLE query. This patch addresses this behavior and adds a specific test for this scenario.

v1.6.5¶

Allow pipes to sync DataFrame generators. If pipe.sync() receives a generator (for DataFrames, dictionaries, or lists), it will attempt to consume it and sync its chunks in parallel threads (this can be single-threaded with --workers 1). For SQL pipes, this will be capped at your configured pool size (default 5) minus the running number of threads.

This means you may now return generators to large transcations, such as reading a large CSV:

def fetch(pipe, **kw) -> Iterable['pd.DataFrame']:
    return pd.read_csv('data.csv', chunksize=1000)

Any iterator of DataFrame-like chunks will work:

def fetch(pipe, **kw) -> Generator[List[Dict[str, Any]]]:
    return (
        [
            {'id': 1, 'val': 10.0 * i},
            {'id': 2, 'val': 20.0 * i},
        ] for i in range(10)
    )

This new behavior has been added to SQLConnector.fetch() so you may now confidently sync very large tables between your databases.

NOTE: The default chunksize for SQL queries has been lowered to 100,000 from 1,000,000. You may alter this value with --chunksize or setting the value in MRSM{system:connectors:sql:chunksize} (you can also edit the default pool size here).

Fix edge case with SQL in-place syncs. Occasionally, a few docs would be duplicated when running in-place SQL syncs. This patch increases the fetch window size to mitigate the issue.
Remove begin and end from filter_existing(). The keyword arguments were interfering with the determined datetime bounds, so this patch removes these flags (albeit begin was already ignored) to avoid confusion. Date bounds are solely determined from the contents of the DataFrame.

v1.6.4¶

Allow for mixed UTC offsets in datetimes. UTC offsets are now applied to datetime values before timezone information is stripped, which should now reflect accurate values. This patch also fixes edge cases when different offsets are synced within the same transcation.

import meerschaum as mrsm
pipe = mrsm.Pipe('a', 'b', columns={'datetime': 'dt'})
pipe.sync([
    {'dt': '2023-01-01 00:00:00+00:00'},
    {'dt': '2023-01-02 00:00:00+01:00'},
])
pipe.get_data().to_dict(orient=records)
# [
#     {'dt': Timestamp('2023-01-01 00:00:00')},
#     {'dt': Timestamp('2023-01-01 23:00:00')}
# ]

Allow skipping datetime detection. The automatic datetime detection feature now respects a pipe's dtypes; columns that aren't of type datetime64[ns] will be ignored.

import meerschaum as mrsm
pipe = mrsm.Pipe('a', 'b', dtypes={'not-date': 'str'})
pipe.sync([
    {'not-date': '2023-01-01 00:00:00'}
])
pipe.get_data().to_dict(orient=records)
# [
#     {'not-date': '2023-01-01 00:00:00'}
# ]

Added utility method enforce_dtypes(). The DataFrame data type enforcement logic of pipe.enforce_dtypes() has been exposed as meerschaum.utils.misc.enforce_dtypes():

from meerschaum.utils.misc import enforce_dtypes
import pandas as pd
df = pd.DataFrame([{'a': '1'}])
enforce_dtypes(df, {'a': 'Int64'}).dtypes
# a    Int64
# dtype: object

Performance improvements. Some of the unnecessarily immutable transformations have been replaced with more memory- and compute-efficient in-place operations. Other small improvements like better caching should also speed things up.
Removed noise from debug output. The virtual environment debug messages have been removed to make --debug easier to read.
Better handle inferred datetime index. The inferred datetime index feature may now be disabled by setting datetime to None. Improvements were made to be handle incorrectly identified indices.
Improve dynamic dtypes for SQLite. SQLite doesn't allow for modifying column types but is usually dynamic with data types. A few edge cases have been solved with a workaround for altering the table's definition.

v1.6.3¶

Fixed an issue with background jobs. A change had broken daemon functionality has been reverted.

v1.6.2¶

Virtual environment and pip tweaks. With upcoming changes to pip coming due to PEP 668, this patch sets the environment variable PIP_BREAK_SYSTEM_PACKAGES when executing pip internally. All packages are installed within virtual environments except uvicorn, gunicorn, and those explicitly installed with a venv of None.
Change how pipes are pretty-printed. Printing the attributes of a single pipe now highlights the keys in blue.
Fix an issue with bootstrap pipes and plugins. When bootstrapping a pipe with a plugin connector, the plugin's virtual environment will now be activated while executing its register() function.
Update dependencies. The minimum version of duckdb was bumped to 0.7.1, duckdb-engine was bumped to 0.7.0, and pip was lowered to 22.0.4 to accept older versions. Additionally, pandas==2.0.0rc1 was tested and confirmed to work, so version 1.7.x of Meerschaum will likely require 2.0+ of pandas to make use of its PyArrow backend.

v1.6.0 – v1.6.1¶

Breaking Changes

Dropped Python 3.7 support. The latest pandas requires 3.8+, so to use Pandas 1.5.x, we have to finally drop Python 3.7.
Upgrade SQLAlchemy to 2.0.5+. This includes better transaction handling with connections. Other packages which use SQLAlchemy may not yet support 2.0+.
Removed MQTTConnector. This was one of the original connectors but was never tested or used in production. It may be reintroduced via a future mqtt plugin.

Bugfixes and Improvements

Stop execution when improper command-line arguments are passed in. Incorrect command-line arguments will now return an error. The previous behavior was to strip the flags and execute the action anyway, which was undesirable.

$ mrsm show pipes -c

 💢 Invalid arguments:
  show pipes -c

   🛑 argument -c/-C/--connector-keys: expected at least one argument

Allow bootstrap connector to create custom connectors. The bootstrap connector wizard can now handle registering custom connectors. It uses the REQUIRED_ATTRIBUTES list set in the custom connector class when determining what to ask for.
Allow custom connectors to omit __init__() If a connector is created via @make_connector and doesn't have an __init__() function, the base one is used to create the connector with the correct type (derived from the class name) and verify the REQUIRED_ATTRIBUTES values if present.
Infer a connector's type from its class name. The type of a connector is now determined from its class name (e.g. FooConnector would have a type foo). When inheriting from Connector, it is no longer required to explictly pass the type before the label. For backwards compatability, the legacy method still behaves as expected.

from meerschaum.connectors import (
    Connector,
    make_connector,
    get_connector,
)

@make_connector
class FooConnector:
    REQUIRED_ATTRIBUTES = ['username', 'password']

conn = get_connector(
    'foo',
    username = 'abc',
    password = 'def',
)

Allow connectors to omit a label. The default label main will be used if label is omitted.
Add meta keys to connectors. Like pipes, the meta property of a connector returns a dictionary with the kwargs needed to reconstruct the connector.

conn = mrsm.get_connector('sql:temp', flavor='sqlite', database=':memory:')
print(conn.meta)
# {'type': 'sql', 'label': 'temp', 'database': ':memory:', 'flavor': 'sqlite'}

Remove NUL bytes when inserting into PostgreSQL. PostgreSQL doesn't support NUL bytes in text ('\0'), so these characters are removed from strings when copying into a table.
Cache pipe.exists() for 5 seconds. Repeated calls to pipe.exists() will be sped up due to short-term caching. This cache is invalidated when syncing or dropping a pipe.
Fix an edge case with subprocesses in headless environments. Checks were added to subprocesses to prevent using interactive features when no such features may be available (i.e. termios).
Added pprint(), get_config(), and attempt_import() to the top-level namespace. Frequently used functions pprint(), get_config(), and attempt_import() have been promoted to the root level of the meerschaum namespace, i.e.:

import meerschaum as mrsm
mrsm.pprint(mrsm.get_config('meerschaum'))

sqlalchemy = mrsm.attempt_import('sqlalchemy')

Fix CLI for MSSQL. The interactive CLI has been fixed for Microsoft SQL Server.

1.5.x Releases¶

Read release notes

The 1.5.x series offered many great improvements, namely the ability to use an integer datetime axis and the addition of JSON columns.

v1.5.8 – v1.5.10¶

Infer JSON columns from the first first non-null value. When determining complex columns (dictionaries or lists), the first non-null value of the dataframe is checked rather than the first row only. This accounts for documents which contain variable keys in the same sync, e.g.:

import meerschaum as mrsm
pipe = mrsm.Pipe('a', 'b')
pipe.sync([
    {'a': {'b': 1}},
    {'c': {'d': 2}},
])

Fix a bug when reconstructing JSON columns. When rebuilding JSON values after merging, a check is first performed if the value is in fact a string (sometimes NULLS slip in).
Increase the timeout when determining Python versions. This fixes some difficult-to-reproduce bugs on Windows.

v1.5.7¶

Replace ast.literal_eval() with json.loads() when filtering JSON columns. This patch replaces the use of str and ast.literal_eval() with json.dumps() and json.loads() to preserve accuracy.
Fix a subtle bug with subprocesses. The function run_python_package() now better handles environment passing and raises a more verbose warning when something goes wrong.
Allow columns with 'create' in the name. A security measure previously disallowed certain keywords when sanitizing input. Now columns are allowed to contain certain keywords.

v1.5.3 – v1.5.6¶

Pipes now support syncing dictionaries and lists. Complex columns (dicts or lists) will now be preserved:

import meerschaum as mrsm
pipe = mrsm.Pipe('a', 'b')
pipe.sync([{'a': {'b': 1}}])
df = pipe.get_data()
print(df['a'][0])
# {'b': 1}

You can also force strings to be parsed by setting the data type to json:

import meerschaum as mrsm
pipe = mrsm.Pipe(
    'foo', 'bar',
    columns = {'datetime': 'id'},
    dtypes = {'data': 'json', 'id': 'Int64'},
)
docs = [{'id': 1, 'data': '{"foo": "bar"}'}]
pipe.sync(docs)
df = pipe.get_data()
print(df['data'][0])
# {'foo': 'bar'}

For PostgreSQL-like databases (e.g. TimescaleDB), this is stored as JSONB under the hood. For all others, it's stored as the equivalent for TEXT.

Fixed determining the version when installing plugins. Like the required list, the __version__ string must be explicitly set in order for the correct version to be determined.
Automatically cast postgres to postgresql When a SQLConnector is built with a flavor of postgres, it will be automatically set to postgresql.

v1.5.0 – v1.5.2¶

Pipes may now use integers for the datetime column. If you use an auto-incrementing integer as your primary key, you may now use that column as your pipe's datetime column, just specify the dtype as an Int64:

import meerschaum as mrsm
pipe = mrsm.Pipe(
    'foo', 'bar',
    instance = 'sql:memory',
    columns = {
        'datetime': 'id',
    },
    dtypes = {
        'id': 'Int64',
    },
)
pipe.sync([{'id': 1, 'foo': 'bar'}])
pipe.sync([{'id': 2, 'foo': 'baz'}])

This applies the same incremental range filtering logic as is normally done on the datetime axis.

Allow for multiple plugins directories. You may now set multiple directories for MRSM_PLUGINS_DIR. All of the plugins contained in each directory will be symlinked together into a single plugins namespace. To do this, just set MRSM_PLUGINS_DIR to a JSON-encoded list:

export MRSM_PLUGINS_DIR='["./plugins_1", "./plugins_2"]'

Better Windows support. At long last, the color issues plaguing Windows users have finally been resolved. Additionally, support for background jobs has been fixed on Windows, though the daemonization library I use is pretty hacky and doesn't make for the smoothest experience. But at least it works now!
Fixed unsafe TAR extraction. A PR about unsafe use of tar.extractall() brought this issue to light.
Fixed the blank logs bug in show logs. Backtracking a couple lines before following the rest of the logs has been fixed.
Requirements may include brackets. Python packages listed in a plugin's requirements list may now include brackets (e.g. meerschaum[api]).
Enforce 1000 row limit in SQLConnector.to_sql() for SQLite. When inserting rows, the chunksize of 1000 is enforced for SQLite (was previously enforced only for reading).
Patch parameters from --params in edit pipes and register pipes. When editing or registering pipes, the value of --params will now be patched into the pipe's parameters. This should be very helpful when scripting.
Fixed edit users. This really should have been fixed a long time ago. The action edit users was broken due to a stray import left over from a major refactor.
Fixed a regex bug when cleaning up packages.
Removed show gui and show modules.

1.4.x Releases¶

Read release notes

The 1.4.x series brought some incredible, stable releases, and the highlight feature was in-place SQL syncs for massive performance improvement. The addition of temporary to Pipes also made using pipes in projects more accessible.

v1.4.14¶

Added flag temporary to Pipe (and --temporary). Pipes built with temporary=True, will not create instance tables (pipes, users, and plugins) or be able to modify registration. This is particularly useful when creating pipes from existing tables when automatic registration is not desired.

import meerschaum as mrsm
import pandas as pd
conn = mrsm.get_connector('sql:temp', uri='postgresql://user:pass@localhost:5432/db')

### Simulating an existing table.
table_name = 'my_table'
conn.to_sql(
    pd.DataFrame([{'id_column': 1, 'value': 1.0}]),
    name = table_name,
)

### Create a temporary pipe with the existing table as its target.
pipe = mrsm.Pipe(
    'foo', 'bar',
    target = table_name,
    temporary = True,
    instance = conn,
    columns = {
        'id': 'id_column',
    },
)

docs = [
    {
        "id_column": 1,
        "value": 123.456,
        "new_column": "hello, world!",
    },
]

### Existing table `my_table` is synced without creating other tables
### or affecting pipes' registration.
pipe.sync(docs)

Fixed potential security of public instance tables. The API now refuses to sync or serve data if the target is a protected instance table (pipes, users, or plugins).
Added not-null check to pipe.get_sync_time(). The datetime column should never contain null values, but just in case, pipe.get_sync_time() now passes a not-null check to params for the datetime column.
Removed prompt for value from pipe.bootstrap(). The prompt for an optional value column has been removed from the bootstrapping wizard because pipe.columns is now largely used as a collection of indices rather than the original purpose of meta-columns.
Pass --debug and other flags in copy pipes. Command line flags are now passed to the new pipe when copying an existing pipe.

v1.4.12 – v1.4.13¶

Fixed an issue when syncing empty DataFrames (#95). When syncing an empty list of documents, Pipe.filter_existing() would trigger pulling the entire table into memory. This patch adds a check if the dataframe is empty.
Allow the datetime column to be omitted in the bootstrap wizard. Now that the datetime index is optional, the bootstrapping wizard allows users to skip this index.
Fixed a small issue when syncing to MySQL. Due to the addition of MySQL 5.7 support in v1.4.11, a slight edge case arose which broke SQL definitions. This patch fixes MySQL behavior when a WHERE clause is present in the definition.

v1.4.11¶

Add support for older versions of MySQL. The WITH keyword for CTE blocks was not introduced until MySQL 8.0. This patch uses the older syntax for older versions of MySQL and MariaDB. MySQL 5.7 was added to the test suite.
Allow for any iterable in items_str() If an iterable other than a list is passed to items_str(), it will convert to a list before building the string:

from meerschaum.utils.misc import items_str
print(items_str({'apples': 1, 'bananas': 2}, quotes=False)
# apples and bananas

Fixed an edge case with datetime set to None. This patch will ignore the datetime index even if it was set explicitly to None.
Added Pipe.children. To complement Pipe.parents, setting the parameters key children to a list of pipes' keys will be treated the same as Pipe.parents:

import meerschaum as mrsm
pipe = mrsm.Pipe(
    'a', 'b',
    parameters = {
        'children': [
            {
                'connector': 'a',
                'metric': 'b',
                'location': 'c',
            },
        ]
    }
)
print(pipe.children)
# [Pipe('a', 'b', 'c')]

Added support for type:label syntax in mrsm.get_connector(). The factory function mrsm.get_connector() expects the type and label as two arguments, but this patch allows for passing a single string with both arguments:

import meerschaum as mrsm
print(mrsm.get_connector('sql:local'))
# sql:local

Fixed more edge case bugs. For example, converting to Int64 sometimes breaks with older versions of pandas. This patch adds a workaround.

v1.4.10¶

Fixed an issue with syncing background jobs. The --name flag of background jobs with colliding with the name keyword argument of SQLConnector.to_sql().
Fixed a datetime bounding issue when datetime index is omitted. If the minimum datetime value of the incoming dataframe cannot be determined, do not bound the get_data() request.
Keep existing parameters when registering plugin pipes. When a pipe is registered with a plugin as its connector, the return value of the register() function will be patched with the existing in-memory parameters.
Fixed a data type syncing issue. In cases where fetched data types do not match the data types in the pipe's table (e.g. automatic datetime columns), a bug has been patched to ensure the correct data types are enforced.
Added Venv to the root namespace. Now you can access virtual environments directly from mrsm:

import meerschaum as mrsm

with mrsm.Venv('noaa'):
    import pandas as pd

v1.4.9¶

Fixed in-place syncs for aggregate queries. In-place SQL syncs which use aggregation functions are now handled correctly. This version addresses differences in column types between backtrack and new data. For example, the following query will now be correctly synced:

WITH days_src AS (
  SELECT *, DATE_TRUNC('day', "datetime") AS days
  FROM plugin_stress_test
)
SELECT days, AVG(val) AS avg_value
FROM days_src
GROUP BY days

Activate virtual environments for custom instance connectors. All pipe methods now activate virtual environments for custom instance connectors.
Improved database connection performance. Cold connections to a SQL database have been sped up by replacing sqlalchemy_utils with handwritten logic (JSON for PostgreSQL-like and SQLite).
Fixed an issue with virtual environment verification in a portable environment. The portable build has been updated to Python 3.9.15, and this patch includes a check to determine the known site-package path for a virtual environment of None instead of relying on the default user site-packages directory.
Fixed some environment warnings when starting the API

v1.4.5 – v1.4.8¶

Bugfixes and stability improvements. These versions included several bugfixes, such as patching --skip-check-existing for in-place syncs and fixing the behavior of --params (build_where()).

v1.4.0 – v1.4.4¶

Added in-place syncing for SQL pipes. This feature is big (enough to warrant a new point release). When pipes with the same instance connector and data source connector are synced, the method sync_pipe_inplace() is invoked. For SQL pipes, this means the entire syncing process will now happen entirely in SQL, which can lead to massive performance improvements.

import meerschaum as mrsm
import pandas as pd

conn = mrsm.get_connector('sql', 'local')
conn.to_sql(pd.DataFrame([{'a': 1}]), 'foo')

pipe = mrsm.Pipe(
    "sql:local", "foo",
    instance = "sql:local",
    parameters = {
        "query": "SELECT * FROM foo"
    },
)
### This will no longer load table data into memory.
pipe.sync()

This applies even when the source table's schema changes, just like the dynamic columns feature added in v1.3.0.

To disable this behavior, run the command edit config system and set the value under the keys experimental:inplace_sync to false.

Added negation to --params. The build_where() function now allows you to negate certain values when prefixed with an underscore (_):

### Show recent data, excluding where `id` equals `2`.
mrsm show data --params id:_2

Added --params to SQL pipes' queries. Specifying parameters when syncing SQL pipes will add those constraints to the fetch stage.
Skip invalid parameters in --params. If a column does not exist in a pipe's table, the value will be ignored in --params.
Fixed environment issue when starting the Web API with gunicorn.
Added an emoji to the SQL Query option of the web console.
Fixed an edge case with data type enforcement.
Other bugfixes

1.3.x Releases¶

Read release notes

The 1.3.x series brought a tremendous amount of new features and stability improvements. Read below to see everything that was introduced!

v1.3.13¶

Fixed an issue when displaying backtrack data on the Web Console. Certain values like pd.NA would break the Recent Data view on the Web Console. Now the values are cast to strings before building the table.
Added YAML and JSON support to editing parameters. YAML is now the default, and toggle buttons have been added to switch the encoding. Line numbers have also been added to the editors.
Removed the index column from the CSV downloader. When the download button is clicked, the dataframe's index column will be omitted from the CSV file.
Changed the download filename to <target>.csv. The download process will now name the CSV file after the table rather than the pipe.
Web Console improvements. The items in the actions builder now are presented with a monospace font. Actions and subactions will have underscores represented as spaces.
Activating the virtual environment None will not override your current working directory. This is especially useful when testing the API. Activating the virtual environment None will insert behind your current working directory or '' in sys.path.
Added WebSocket Secure Support. This has been coming a long time, so I'm proud to announce that the web console can now detect whether the client is connecting via HTTPS and (assuming the server has the appropriate proxy configuration) will connect via WSS.

v1.3.10 – v1.3.12¶

Fixed virtual environment issues when syncing. This one's a doozy. Before this patch, there would frequently be warnings and sometimes exceptions thrown when syncing a lot of pipes with a lot of threads. This kind of race condition can be hard to pin down, so this patch reworks the virtual environment resolution system by keeping track of which threads have activated the environments and refusing to deactivate if other threads still depend on the environment. To enforce this behavior, most manual ivocations of activate_venv() were replaced with the Venv context manager. Finally, the last stage of each action is to clean up any stray virtual environments. Note: You may still run into the wrong version of a package being imported into your plugin if you're syncing a lot of plugins concurrently.
Allow custom instance connectors to be selected on the web console. Provided all of the appropriate interface methods are implemented, selecting a custom instance connector from the instance dropdown should no longer throw an error.
Bugfixes and improvements to the virtual environment system. This patch should resolve your virtual environment woes but at a somewhat significant performance penalty. Oh well, 'tis the price we must pay for correct and determinstic code!
Fixed custom flags added by add_plugin_argument(). Refactoring work to decouple plugins from the argument parser had the unintended side effect of skipping over custom flags until after sysargs had already been parsed. This patch ensures all plugins with add_plugin_argument in their root module will be loaded before parsing.
Upgraded dash-extensions, dash, and dash-bootstrap-components. At long last, dash-extensions will no longer need to be held back.
Added additional websockets endpoints. The endpoints /ws, /dash/ws/, and /dashws resolve to the same handler. This is to allow compatability with different versions of dash-extensions.
Allow for custom arguments to be added from outside plugins. The function add_plugin_argument() will now accept arguments when executed from outside a plugin.
Fixed verify packages.

v1.3.6 – v1.3.9¶

Allow for syncing multiple data types per column. The highlight of this release is support for syncing multiple data types per column. When different data types are encountered, the underlying column will be converted to TEXT:

import meerschaum as mrsm
pipe = mrsm.Pipe('a', 'b', instance='sql:memory')

pipe.sync([{'a': 1}])
print(pipe.get_data())
#    a
# 0  1

pipe.sync([{'a': 'foo'}])
#      a
# 0    1
# 1  foo

Cleaned up the Web Console. The web console's navbar is now more mobile-friendly, and a "sign out" button has been added.
Removed plugins import when checking for environment connectors. This should make some commands feel more snappy by lazy loading custom connectors.
Fixed an issue with an updated versin of uvicorn.
Fixed an issue with docker-compose.
Fixed an issue with FastAPI on Python 3.7.
Added support for Python 3.11.
Renamed meerschaum.actions.arguments to meerschaum._internal.arguments.

v1.3.4 – v1.3.5¶

Define environment connectors with JSON or URIs. Connectors defined as environment variables may now have their attributes set as JSON in addition to a URI string.
Custom Connectors may now be defined as environment variables. You may now set environment variables for custom connectors defined via @make_connector, e.g.:

MRSM_FOO_BAR='{"foo": "bar"}' mrsm show connectors

Allow for custom connectors to be instance connectors. Add the property IS_INSTANCE = True your custom connector to add it to the official list of instance types:

from meerschaum.connectors import make_connector, Connector

@make_connector
class FooConnector(Connector):
    IS_INSTANCE = True

    def __init__(label: str, **kw):
        super().__init__('foo', label, **kw)

Install packages for all plugins with mrsm install required. The default behavior for mrsm install required with no plugins named is now to install dependencies for all plugins.
Syncing bugfixes.

v1.3.2 – v1.3.3¶

Fixed a bug with begin and end bounds in Pipe.get_data(). A safety measure was incorrectly checking if the quoted version of a column was in pipe.get_columns_types(), not the unquoted version. This patch restores functionality for pipe.get_data().
Fixed an issue with an upgraded version of SQLAlchemy.
Added a parameters editor the Web UI. You may now edit your pipes' parameters in the browser through the Web UI!
Added a SQL query editor to the Web UI. Like the parameters editor, you can edit your pipes' SQL queries in the browser.
Added a Sync Documents option to the Web UI. You can directly sync documents into pipes on the Web UI.
Added the arguments order, limit, begin_add_minutes, and end_add_minutes to Pipe.get_data(). These new arguments will give you finer control over the data selection behavior.
Enforce consistent ordering of indices in Pipe.get_data().
Allow syncing JSON-encoded strings. This patch allows pipes to sync JSON strings without first needing them to be deserialized.
Fixed an environment error with Ubuntu 18.04.
Bumped duckdb and duckdb-engine.
Added a basic CLI for duckdb. This will probably be replaced later down the line.

v1.3.1¶

Fixed data type enforcement issues. A serious bug in data type enforcement has been patched.
Allow Pipe.dtypes to be edited. You can now set keys in Pipe.dtypes and persist them with Pipe.edit().
Added Pipe.update(). Pipe.update() is an alias to Pipe.edit(interactive=False).
Pipe.delete() no longer deletes local attributes. It still removes Pipe.id, but local attributes will now remain intact.
Fixed dynamic columns on DuckDB. DuckDB does not allow for altering tables when indices are created, so this patch will drop and rebuild indices when tables are altered.
Replaced CLOB with NVARCHAR(2000) on Oracle SQL. This may require migrating existing pipes to use the new data type.
Enforce integers are of type INTEGER on Oracle SQL. Lots of data type enforcement has been added for Oracle SQL.
Removed datetime warnings when syncing pipes without a datetime column.
Removed grabbing the current time for the sync time if a sync time cannot be determined.

v1.3.0: Dynamic Columns¶

Improvements

Syncing now handles dynamic columns. Syncing a pipe with new columns will trigger an ALTER TABLE query to append the columns to your table:

import meerschaum as mrsm
pipe = mrsm.Pipe('foo', 'bar', instance='sql:memory')

pipe.sync([{'a': 1}])
print(pipe.get_data())
#    a
# 0  1

pipe.sync([{'b': 1}])
print(pipe.get_data())
#       a     b
# 0     1  <NA>
# 1  <NA>     1

If you've specified index columns, you can use this feature to fill in NULL values in your table:

import meerschaum as mrsm
pipe = mrsm.Pipe(
    'foo', 'bar',
    columns = {'id': 'id_col'},
    instance = 'sql:memory',
)

pipe.sync([{'id_col': 1, 'a': 10.0}])
pipe.sync([{'id_col': 1, 'b': 20.0}])

print(pipe.get_data())
#    id_col     a     b
# 0       1  10.0  20.0

Add as many indices as you like. In addition to the special index column labels datetime, id, and value, the values of all keys within the Pipe.columns dictionary will be treated as indices when creating and updating tables:

import meerschaum as mrsm
indices = {'micro': 'station', 'macro': 'country'}
pipe = mrsm.Pipe('demo', 'weather', columns=indices, instance='sql:memory')

docs = [{'station': 1, 'country': 'USA', 'temp_f': 80.6}]
pipe.sync(docs)

docs = [{'station': 1, 'country': 'USA', 'temp_c': 27.0}]
pipe.sync(docs)

print(pipe.get_data())
#    station  country  temp_f  temp_c
# 0        1      USA    80.6    27.0

Added a default 60-second timeout for pipe attributes. All parameter properties (e.g. Pipe.columns, Pipe.target, Pipe.dtypes, etc.) will sync with the instance every 60 seconds. The in-memory attributes will be patched on top of the database values, so your unsaved state won't be lost (persist your state with Pipe.edit()). You can change the timeout duration with mrsm edit config pipes under the keys attributes:local_cache_timeout_seconds. To disable this caching behavior, set the value to null.
Added custom indices and Pandas data types to the Web UI.

Breaking Changes

Removed None as default for uninitalized properties for pipes. Parameter properties like Pipe.columns, Pipe.parameters, etc. will now always return a dictionary, even if a pipe is not registered.
Pipe.get_columns() now sets error to False by default. Pipes are now mostly index-agnostic, so these checks are no longer needed. This downgrades errors in several functions to just warnings, e.g. Pipe.get_sync_time().

Bugfixes

Always quote tables that begin with underscores in Oracle.
Always refresh the metadata when grabbing sqlalchemy tables for pipes to account for dynamic state.

1.2.x Releases¶

Read release notes

This series brought many industry-ready features, such as the @make_connector decorator, improvements to the virtual environment system, the environment variable MRSM_PLUGINS_DIR, and much more.

v1.2.9¶

Added support for Windows junctions for virtual environments. This included many changes to fix functionality on Windows. For example, the addition of the MRSM_PLUGINS_DIR environment variable broke Meerschaum on Windows, because Windows requires administrator rights to create symlinks.

v1.2.8¶

Custom connectors may now have register(pipe) methods. Just like the module-level register(pipe) plugin function, custom connectors may also provide this function as a class member.
Print a traceback if fetch(pipe) breaks. A more verbose traceback is printed if a plugin breaks during the syncing process.
Cleaned up sync pipes output. This patch cleans up the syncing process's pretty output.
Respect --nopretty in sync pipes. This flag will only print JSON-encoded dictionaries for sync pipes. Tracebacks may still interfere without standard output, however.

v1.2.5 – v1.2.7¶

Venv context managers do not deactivate previously activated venvs. You can safely use Venv without worrying about deactivating your previously activated environments.
Better handling of nested plugin dependencies.
Plugin.get_dependencies() will not trigger an import. If you want certainty about a plugin's required list, trigger an import manually. Otherwise, it will use ast.literal_eval() to determine the required list from the source itself. This only works for statically set required lists.
Provide rich traceback for broken plugins. If a plugin fails to import, a nice traceback is printed out in addition to a warning.
Only cache Pipe.dtypes if the pipe exists.
Pass current environment to subprocesses. This should retain any custom configuration you've set in the main process.
Hard-code port 5432 as the target DB container port in the stack. Changing the host port now will not change the target port in the container.
Fixed a bug with background jobs and to_sql(). The --name flag was conflicting with to_sql().
Reimplemented apply_patch_to_config(). This patch removes cascadict as a vendored dependency and replaces it with a simpler implementation.
Removed network request for shell connectivity status. The shell now simply checks for the existence of the connector. This may occasionally print an inaccurate connection status, but the speed benefit is worth it.
Moved dill and other required dependencies into the sql dependency group.
Replaced redengine with rocketry.
Patched Literal into typing for Python 3.7.
Fixed shell commands. This includes falling back to ' '.join instead of shlex.join for Python 3.7.

v1.2.1 – v1.2.4¶

Added the @make_connector decorator. Plugins may now extend the base Connector class to provide custom connectors. For most cases, the built-in plugin connector should work fine. This addition opens up the internal connector system so that plugin authors may now add new types. See below for a minimal example of a new connector class:

# ~/.config/meerschaum/plugins/example.py

from meerschaum.connectors import make_connector, Connector

REQUIRED_ATTRIBUTES = {'username', 'password'}

@make_connector
class FooConnector(Connector):

    def __init__(self, label: str, **kw):
        """
        Instantiate the base class and verify the required attributes are present.
        """
        kw['label'] = label
        super().__init__('foo', **kw)
        self.verify_attributes(REQUIRED_ATTRIBUTES)


    def fetch(self, pipe, **kw):
        """
        Return a dataframe (or list of dicts / dict of lists).
        The same as you would in a regular fetch plugin, except that now
        we can store configuration within the connector itself.
        """
        return [{self.username: '2020-01-01'}]

Allow for omitting datetime column index. The datetime column name is still highly recommended, but recent changes have allowed pipes to be synced without a dedicated datetime axis. Plenty of warnings will be thrown if you sync your pipes without specifying a datetime index. If a datetime column can be found, it will be used as the index.

import meerschaum as mrsm
### This will work but throw warnings.
mrsm.Pipe('foo', 'bar').sync([{'a': 1}])

v1.2.0¶

Improvements

Added the action start connectors. This command allows you to wait until all of the specified connectors are available and accepting connections. This feature is very handy when paired with the new MRSM_SQL_X URI environment variables.
Added MRSM_PLUGINS_DIR. This one's been on my to-do list for quite a bit! You can now place your plugins in a dedicated, version-controlled directory outside of your root directory.

Like MRSM_ROOT_DIR, specify a path with the environment variable MRSM_PLUGINS_DIR:

MRSM_PLUGINS_DIR=plugins \
  mrsm show plugins

Allow for symlinking in URI environment variables. You may now reference configuration keys within URI variables:

MRSM_SQL_FOO=postgresql://user:MRSM{meerschaum:connectors:sql:main:password}@localhost:5432/db

Increased token expiration window to 12 hours. This should reduce the number of login requests needed.
Improved virtual environment verification. More edge cases have been addressed.
Symlink Meerschaum into Plugin virtual environments. If plugins do not specify Meerschaum in the required list, Meerschaum will be symlinked to the currently running package.

Breaking changes

API endpoints for registering and editing users changed. To comply with OAuth2 convention, the API endpoint for registering a user is now a url-encoded form submission to /users/register (/user/edit for editing).

You must upgrade both the server and client to v1.2.0+ to login to your API instances.
Replaced meerschaum.utils.sql.update_query() with meerschaum.utils.sql.get_update_queries(). The new function returns a list of query strings rather than a single query. These queries are executed within a single transaction.

Bugfixes

Removed version enforcement in pip_install(). This changed behavior allows for custom version constraints to be specified in Meerschaum plugins.
Backported UPDATE FROM query for older versions of SQLite. The current mutable data logic uses an UPDATE FROM query, but this syntax is only present in versions of SQLite greater than 3.33.0 (released 2020-08-14). This releases splits the same logic into DELETE and INSERT queries for older versions of SQLite.
Fixed missing suggestions for shell-only commands. Completions for commands like instance are now suggested.
Fixed an issue with killing background jobs. The signals were not being sent correctly, so this release includes better job process management.

1.1.x Releases¶

Read release notes

The 1.1.x series brought a lot of great new features, notably connector URI parsing (e.g. MRSM_SQL_<LABEL>), parsing underscores as spaces in actions, and rewriting the Docker image to run at as a normal user.

v1.1.9 – v1.1.10¶

Fixed plugins virtual environments. A typo in v1.1.8 temporarily broke plugins, and this patch fixes that change.
Fixed Meerschaum on Windows. A change in a previous release allowed for dist-packages for the root user (not advised but supported). The check for root (os.geteuid()) does not exist on Windows, so this patch accounts for that behavior.
Tweaked screen clearing on Windows. Meerschaum now always uses clear or cls on Windows instead of ANSI escape sequences.

v1.1.5 – v1.1.8¶

Fixed MRSM_PATCH behavior. In the docker image, MRSM_PATCH is used to overwrite host for sql:main. This patch restores that behavior (with a performance boost).
Fixed virtual environment verification. This patch prevents circular symlinks.
Fixed manually_import_module(). Previous refactoring efforts had broken manually_import_module().
Refactoring While trying to implement multi-thread configuration patching (discarded for the time being), much of the configuration system was cleaned up.

v1.1.1 – v1.1.4¶

Bugfixes

The first four versions following the initial v1.1.0 release addressed breaking changes and edge cases. Below are some notable issues resolved:

Fixed broken Docker images. Changes to the environment and package systems broke functionality of the Docker images. For example, v1.1.0 switched to a stricter package management policy, but this new policy broke the mechanism behind the Docker images (user-level vs venv-level packages).
Verify virtual environments for multiple Python versions. When a virtual environment is first activated, Meerschaum now verifies that the python symlinks point to the correct versions. This is necessary due to a quirk in venv and virtualenv where activating an existing environment with a different Python version overwrites the existing python symlink. It also ensures that the symlinks specify the correct version number, e.g. python3.10. This bevavior is now automatic but may be invoked with mrsm verify venvs.
Fixed inconsistent environment behavior with gunicorn. This one was tricky to troubleshoot. Due to the migration to the user-level Docker image, a subtle bug surfaced where the environment variables for gunicorn were incorrectly serialized.
Fixed slow-performing edge cases in determine_version(). Inconsistencies in naming conventions in some packages like pygments led to failures to quickly determine the version.
Fixed Web API actions. In v1.1.0, the default virtual environment was pinned to mrsm, and this broke a function which relied on the old inferred default value of None. Always remember: explicit is better than implicit.
Fixed start job for existing jobs. The same naming change broke daemon_action(). Explcit code is important, folks!

v1.1.0¶

What's New

Underscores in actions may now be parsed as spaces. This took way more work than expected, but anyway, custom actions with underscores in the function names are now treated as spaces! Consider the following:

@make_action
def foo_bar(**kw):
    return True, "Huzzah!"

The above action may now be executed as foo bar or foo_bar:

mrsm foo bar

Create a SQLConnector or APIConnector directly from a URI. If you already have a connection string, you can skip providing credentials and build a connector directly from the URI. If you omit a label, then the lowercase form of '<username>@<host>/<database>' is used:

from meerschaum.connectors import SQLConnector
uri = 'postgresql://user:pass@localhost:5432/db'

conn = SQLConnector(uri=uri)
print(conn)
# sql:user@localhost/db

conn = SQLConnector('foo', uri=uri)
print(conn)
# sql:foo

conn = SQLConnector.from_uri(uri)
print(conn)
# sql:user@localhost/db

The APIConnector may also be built from a URI:

from meerschaum.connectors import APIConnector
uri = 'http://user:pass@localhost:8000'

conn = APIConnector(uri=uri)
print(conn)
# api:username@localhost

conn = APIConnector('bar', uri=uri)
print(conn)
# api:bar

conn = APIConnector.from_uri(uri)
print(conn)
# api:user@localhost

Define temporary connectors from URIs in environment variables. If you set environment variables with the format MRSM_SQL_<LABEL> to valid URIs, new connectors will be available under the keys sql:<label>, where <label> is the lowercase form of <LABEL>:

export MRSM_SQL_FOO=sqlite://///path/to/sqlite.db
mrsm show connectors sql:foo

You can set as many connectors as you like, and they're treated the same as connectors registered in your permanent configuration.

### The following environment variable was already exported:
### MRSM_SQL_FOO=sqlite://///path/to/sqlite.db

import meerschaum as mrsm
conn = mrsm.get_connector('sql', 'foo')
print(conn.database)
# /path/to/sqlite.db

Bugfixes

Resolved issues with conflicting virtual and base environments.
Only reinstall a package available at the user-level if its version doesn't match. This was a subtle bug, but now packages are handled strictly in virtual environments except when an appropriate version is available. This may slow down performance, but the change is necessary to ensure a consistent environment.

Potentially Breaking Changes

The database file path for SQLite and DuckDB is now required. When creating a SQLConnector with the flavors sqlite or duckdb, the attribute database (a file path or :memory:) is now required.
Removed --config and --root-dir. These flags were added very early on but have always caused issues. Instead, please use the environment variables MRSM_CONFIG or MRSM_PATCH for modifying the runtime configuration, and use MRSM_ROOT_DIR to specify a file path to the root Meerschaum directory.
Virtual environments must be None for standard library packages. When importing a built-in module with attempt_import(), specify venv=None to avoid attempting installation.
The Docker image now runs as meerschaum instead of root. For improved security, the docker image now runs at a lower privilege.

1.0.x Releases¶

Read release notes

The v1.0.0 release was big news. A ton of features, bugfixes, and perfomance improvements were introduced: for example, v1.0.0 brought support for mutable pipes and data type enforcement. Later releases in the v1.0.x series included --schedule, the Venv context manager, and a whole lot of environment bugfixes.

v1.0.6¶

Plugins may now have requirements.txt. If a plugin contains a file named requirements.txt, the file will be parsed alongside the packages specified in the required list.
Added the module meerschaum.utils.venv. Functions related to virtual environment management have been migrated from meerschaum.utils.packages to meerschaum.utils.venv.
Added the Venv class. You can now manage your virtual environments with the Venv context manager:

from meerschaum.utils.venv import Venv

with Venv():
    import pandas

with Venv('foo'):
    import bar

You can also activate the environments for a Plugin:

from meerschaum.utils.venv import Venv
from meerschaum.plugins import Plugin

with Venv(Plugin('noaa')):
    import requests

Removed --isolated from pip_install. Virtual environments will now respect environment variables and your global pip configuration (~/.pip/pip.conf).
Fixed issues for Python 3.7

v1.0.3 – v1.0.5¶

Fixed environment bugs. This patch resolves issues with the environment variables MRSM_ROOT_DIR, MRSM_CONFIG, and MRSM_PATCH as well as the configuration directories patch_config and permanent_patch_config.
Fixed package management system. Meerschaum better handles package metadata, resolving some annoying issues. See meerschaum.utils.packages.get_module_path() for an example of the improved virtual environment management system. Also, wheel is automatically installed when new packages are installed into new virtual environments.
Set the default venv to 'mrsm'. In all functions declared in meerschaum.utils.packages, the default value of venv is always 'mrsm'. Use None for the venv to use the user's site packages.
Updated dependencies.
Added python-dotenv as a dependency.
Fixed a catalog issue with duckdb.
Updated the testing suite.
More refactoring. Code needs to be beautiful!

v1.0.2¶

Allow id column to be omitted. When generating the UPDATE query, the id column may now be omitted (NOTE: the datetime column will be assumed to be the primary key in this scenario).
Added --schedule (-s or --cron). The --schedule flag (-s) now lets you schedule any command to be executed regulary, not unlike crontab. This can come in handy with --daemon (-d), e.g.:

mrsm sync pipes -c plugin:foo -s hourly -d

Here is more information on the scheduling syntax.

Fixed an issue with SQLite. An issue where the value columns not loading in SQLite has been addressed.

v1.0.1¶

Added citus as an official database flavor. Citus is a distributed database built on PostgreSQL. When an id column is provided, Meerschaum will call create_distributed_table() on the pipe's ID index. Citus has also been added to the official test suite.
Changed end datetimes to be exclusive. The end parameter now generates < instead of <=. This shouldn't be a major breaking change but is important to be aware of.
Bumped rich to v12.4.4.

v1.0.0: Mutable at Last¶

What's New

Inserts and Updates An additional layer of processing separates new rows from updated rows. Meerschaum uses your datetime and id columns (if you specified an id column) to determine which rows have changed. Therefore a primary key is not required, as long as the datetime column is unique or the datetime and id columns together emulate a composite primary key.

Meerschaum will insert new rows as before as well as creating a temporary table (same name as the pipe's target but with a leading underscore). The syncing engine then issues the appropriate MERGE or UPDATE query to update all of the rows in a batch.

For example, the following lines of code will result in a table with only 1 row:
```
>>> import meerschaum as mrsm
>>> pipe = mrsm.Pipe('foo', 'bar', columns={'datetime': 'dt', 'id': 'id'})
>>>
>>> ### Insert the first row.
>>> pipe.sync([{'dt': '2022-06-26', 'id': 1, 'value': 10}])
>>>
>>> ### Duplicate row, no change.
>>> pipe.sync([{'dt': '2022-06-26', 'id': 1, 'value': 10}])
>>>
>>> ### Update the value columns of the first row.
>>> pipe.sync([{'dt': '2022-06-26', 'id': 1, 'value': 100}])
```
Data Type Enforcement. Incoming DataFrames will be cast to the pipe's existing data types, and if you want total control, you can manually specify the Pandas data types for as many columns as you like under the dtypes key of Pipe.parameters, e.g.:
```
columns:
  datetime: timestamp_utc
  id: station_id
dtypes:
  timestamp_utc: 'datetime64[ns]'
  station_id: 'Int64'
  station_name: 'object'
```
Allow for NULL in INT columns. Before pandas v1.3.0, including a null value in an int column would cast it to a float. Now pd.NA has been added and is leveraged in Meerschaum's data type inference system.
Plugins respect .gitignore When publishing a plugin that is contained in a git repository, Meerschaum will parse your .gitignore file to determine which files to omit.
Private API Mode. Require authentication on all API endpoints with start api --private.
No Authentication API Mode. Adding --no-auth will disable all authentication on all API endpoints, including the web console.

Bugfixes

Plugin packaging A breaking bug in the process of packaging and publishing Meerschaum plugins has been patched.
Correct object names in Oracle SQL. Oracle has finally been brought up to speed with other flavors.
Multi-module plugins fix. A small but important fix for multi-module plugins has been applied for their virtual environments.
Improved virtual environment handling for Debian systems. If venv is not available, Meerschaum now better handles falling back to virtualenv.
Allow for syncing lists of dicts. In addition to syncing a dict of lists, Pipe.sync() now supports a list of dicts.
Allow for begin to equal None for Pipe.fetch(). The behavior of determining begin from Pipe.get_sync_time() only takes place when begin is omitted, not when it is None. Now begin=None will not add a lower bound to the query.