mox/vendor/github.com/mjl-/bstore/doc.go
Mechiel Lukkien 849b4ec9e9
add webmail
it was far down on the roadmap, but implemented earlier, because it's
interesting, and to help prepare for a jmap implementation. for jmap we need to
implement more client-like functionality than with just imap. internal data
structures need to change. jmap has lots of other requirements, so it's already
a big project. by implementing a webmail now, some of the required data
structure changes become clear and can be made now, so the later jmap
implementation can do things similarly to the webmail code. the webmail
frontend and webmail are written together, making their interface/api much
smaller and simpler than jmap.

one of the internal changes is that we now keep track of per-mailbox
total/unread/unseen/deleted message counts and mailbox sizes.  keeping this
data consistent after any change to the stored messages (through the code base)
is tricky, so mox now has a consistency check that verifies the counts are
correct, which runs only during tests, each time an internal account reference
is closed. we have a few more internal "changes" that are propagated for the
webmail frontend (that imap doesn't have a way to propagate on a connection),
like changes to the special-use flags on mailboxes, and used keywords in a
mailbox. more changes that will be required have revealed themselves while
implementing the webmail, and will be implemented next.

the webmail user interface is modeled after the mail clients i use or have
used: thunderbird, macos mail, mutt; and webmails i normally only use for
testing: gmail, proton, yahoo, outlook. a somewhat technical user is assumed,
but still the goal is to make this webmail client easy to use for everyone. the
user interface looks like most other mail clients: a list of mailboxes, a
search bar, a message list view, and message details. there is a top/bottom and
a left/right layout for the list/message view, default is automatic based on
screen size. the panes can be resized by the user. buttons for actions are just
text, not icons. clicking a button briefly shows the shortcut for the action in
the bottom right, helping with learning to operate quickly. any text that is
underdotted has a title attribute that causes more information to be displayed,
e.g. what a button does or a field is about. to highlight potential phishing
attempts, any text (anywhere in the webclient) that switches unicode "blocks"
(a rough approximation to (language) scripts) within a word is underlined
orange. multiple messages can be selected with familiar ui interaction:
clicking while holding control and/or shift keys.  keyboard navigation works
with arrows/page up/down and home/end keys, and also with a few basic vi-like
keys for list/message navigation. we prefer showing the text instead of
html (with inlined images only) version of a message. html messages are shown
in an iframe served from an endpoint with CSP headers to prevent dangerous
resources (scripts, external images) from being loaded. the html is also
sanitized, with javascript removed. a user can choose to load external
resources (e.g. images for tracking purposes).

the frontend is just (strict) typescript, no external frameworks. all
incoming/outgoing data is typechecked, both the api request parameters and
response types, and the data coming in over SSE. the types and checking code
are generated with sherpats, which uses the api definitions generated by
sherpadoc based on the Go code. so types from the backend are automatically
propagated to the frontend.  since there is no framework to automatically
propagate properties and rerender components, changes coming in over the SSE
connection are propagated explicitly with regular function calls.  the ui is
separated into "views", each with a "root" dom element that is added to the
visible document. these views have additional functions for getting changes
propagated, often resulting in the view updating its (internal) ui state (dom).
we keep the frontend compilation simple, it's just a few typescript files that
get compiled (combined and types stripped) into a single js file, no additional
runtime code needed or complicated build processes used.  the webmail is served
is served from a compressed, cachable html file that includes style and the
javascript, currently just over 225kb uncompressed, under 60kb compressed (not
minified, including comments). we include the generated js files in the
repository, to keep Go's easily buildable self-contained binaries.

authentication is basic http, as with the account and admin pages. most data
comes in over one long-term SSE connection to the backend. api requests signal
which mailbox/search/messages are requested over the SSE connection. fetching
individual messages, and making changes, are done through api calls. the
operations are similar to imap, so some code has been moved from package
imapserver to package store. the future jmap implementation will benefit from
these changes too. more functionality will probably be moved to the store
package in the future.

the quickstart enables webmail on the internal listener by default (for new
installs). users can enable it on the public listener if they want to. mox
localserve enables it too. to enable webmail on existing installs, add settings
like the following to the listeners in mox.conf, similar to AccountHTTP(S):

	WebmailHTTP:
		Enabled: true
	WebmailHTTPS:
		Enabled: true

special thanks to liesbeth, gerben, andrii for early user feedback.

there is plenty still to do, see the list at the top of webmail/webmail.ts.
feedback welcome as always.
2023-08-07 21:57:03 +02:00

206 lines
11 KiB
Go

/*
Package bstore is a database library for storing and querying Go values.
Bstore is designed as a small, pure Go library that still provides most of
the common data consistency requirements for modest database use cases. Bstore
aims to make basic use of cgo-based libraries, such as sqlite, unnecessary.
Bstore implements autoincrementing primary keys, indices, default values,
enforcement of nonzero, unique and referential integrity constraints, automatic
schema updates and a query API for combining filters/sorting/limits. Queries
are planned and executed using indices for fast execution where possible.
Bstore is designed with the Go type system in mind: you typically don't have to
write any (un)marshal code for your types.
# Field types
Struct field types currently supported for storing, including pointers to these
types, but not pointers to pointers:
- int (as int32), int8, int16, int32, int64
- uint (as uint32), uint8, uint16, uint32, uint64
- bool, float32, float64, string, []byte
- Maps, with keys and values of any supported type, except keys with pointer types.
- Slices and arrays, with elements of any supported type.
- time.Time
- Types that implement binary.MarshalBinary and binary.UnmarshalBinary, useful
for struct types with state in private fields. Do not change the
(Un)marshalBinary method in an incompatible way without a data migration.
- Structs, with fields of any supported type.
Note: int and uint are stored as int32 and uint32, for compatibility of database
files between 32bit and 64bit systems. Where possible, use explicit (u)int32 or
(u)int64 types.
Cyclic types are supported, but cyclic data is not. Attempting to store cyclic
data will likely result in a stack overflow panic.
Anonymous struct fields are handled by taking in each of the anonymous struct's
fields as a type's own fields. The named embedded type is not part of the type
schema, and with a Query it can currently only be used with UpdateField and
UpdateFields, not for filtering.
Bstore embraces the use of Go zero values. Use zero values, possibly pointers,
where you would use NULL values in SQL.
# Struct tags
The typical Go struct can be stored in the database. The first field of a
struct type is its primary key, must always be unique, and in case of an
integer type the insertion of a zero value automatically changes it to the next
sequence number by default. Additional behaviour can be configured through
struct tag "bstore". The values are comma-separated. Typically one word, but
some have multiple space-separated words:
- "-" ignores the field entirely, not stored.
- "name <fieldname>", use "fieldname" instead of the Go type field name.
- "nonzero", enforces that field values are not the zero value.
- "noauto", only valid for integer types, and only for the primary key. By
default, an integer-typed primary key will automatically get a next value
assigned on insert when it is 0. With noauto inserting a 0 value results in an
error. For primary keys of other types inserting the zero value always results
in an error.
- "index" or "index <field1>+<field2>+<...> [<name>]", adds an index. In the
first form, the index is on the field on which the tag is specified, and the
index name is the same as the field name. In the second form multiple fields can
be specified, and an optional name. The first field must be the field on which
the tag is specified. The field names are +-separated. The default name for the
second form is the same +-separated string but can be set explicitly with the
second parameter. An index can only be set for basic integer types, bools, time
and strings. A field of slice type can also have an index (but not a unique
index, and only one slice field per index), allowing fast lookup of any single
value in the slice with Query.FilterIn. Indices are automatically (re)created
when registering a type. Fields with a pointer type cannot have an index.
String values used in an index cannot contain a \0.
- "unique" or "unique <field1>+<field2>+<...> [<name>]", adds an index as with
"index" and also enforces a unique constraint. For time.Time the timezone is
ignored for the uniqueness check.
- "ref <type>", enforces that the value exists as primary key for "type".
Field types must match exactly, e.g. you cannot reference an int with an int64.
An index is automatically created and maintained for fields with a foreign key,
for efficiently checking that removed records in the referenced type are not in
use. If the field has the zero value, the reference is not checked. If you
require a valid reference, add "nonzero".
- "default <value>", replaces a zero value with the specified value on record
insert. Special value "now" is recognized for time.Time as the current time.
Times are parsed as time.RFC3339 otherwise. Supported types: bool
("true"/"false"), integers, floats, strings. Value is not quoted and no escaping
of special characters, like the comma that separates struct tag words, is
possible. Defaults are also replaced on fields in nested structs, slices
and arrays, but not in maps.
- "typename <name>", override name of the type. The name of the Go type is
used by default. Can only be present on the first field (primary key).
Useful for doing schema updates.
# Schema updates
Before using a Go type, you must register it for use with the open database by
passing a (possibly zero) value of that type to the Open or Register functions.
For each type, a type definition is stored in the database. If a type has an
updated definition since the previous database open, a new type definition is
added to the database automatically and any required modifications are made and
checked: Indexes (re)created, fields added/removed, new
nonzero/unique/reference constraints validated.
As a special case, you can change field types between pointer and non-pointer
types. With one exception: changing from pointer to non-pointer where the type
has a field that must be nonzero is not allowed. The on-disk encoding will not be
changed, and nil pointers will turn into zero values, and zero values into nil
pointers. Also see section Limitations about pointer types.
Because named embed structs are not part of the type definition, you can
wrap/unwrap fields into a embed/anonymous struct field. No new type definition
is created.
Some schema conversions are not allowed. In some cases due to architectural
limitations. In some cases because the constraint checks haven't been
implemented yet, or the parsing code does not yet know how to parse the old
on-disk values into the updated Go types. If you need a conversion that is not
supported, you will need to write a manual conversion, and you would have to
keep track whether the update has been executed.
Changes that are allowed:
- From smaller to larger integer types (same signedness).
- Removal of "noauto" on primary keys (always integer types). This updates the
"next sequence" counter automatically to continue after the current maximum
value.
- Adding/removing/modifying an index, including a unique index. When a unique
index is added, the current records are verified to be unique.
- Adding/removing a reference. When a reference is added, the current records
are verified to be valid references.
- Add/remove a nonzero constraint. Existing records are verified.
Conversions that are not currently allowed, but may be in the future:
- Signedness of integer types. With a one-time check that old values fit in the new
type, this could be allowed in the future.
- Conversions between basic types: strings, []byte, integers, floats, boolean.
Checks would have to be added for some of these conversions. For example,
from string to integer: the on-disk string values would have to be valid
integers.
- Types of primary keys cannot be changed, also not from one integer type to a
wider integer type of same signedness.
# BoltDB and storage
BoltDB is used as underlying storage. BoltDB stores key/values in a single
file, in multiple/nested buckets (namespaces) in a B+tree and provides ACID
transactions. Either a single write transaction or multiple read-only
transactions can be active at a time. Do not start a blocking read-only
transaction while holding a writable transaction or vice versa, this will cause
deadlock.
BoltDB returns Go values that are memory mapped to the database file. This
means BoltDB/bstore database files cannot be transferred between machines with
different endianness. BoltDB uses explicit widths for its types, so files can
be transferred between 32bit and 64bit machines of same endianness. While
BoltDB returns read-only memory mapped byte slices, bstore only ever returns
parsed/copied regular writable Go values that require no special programmer
attention.
For each Go type opened for a database file, bstore ensures a BoltDB bucket
exists with two subbuckets:
- "types", with type descriptions of the stored records. Each time the database
file is opened with a modified Go type (add/removed/modified
field/type/bstore struct tag), a new type description is automatically added,
identified by sequence number.
- "records", containing all data, with the type's primary key as BoltDB key,
and the encoded remaining fields as value. The encoding starts with a
reference to a type description.
For each index, another subbucket is created, its name starting with "index.".
The stored keys consist of the index fields followed by the primary key, and an
empty value.
# Limitations
Bstore has limitations, not all of which are architectural so may be fixed in
the future.
Bstore does not implement the equivalent of SQL joins, aggregates, and many
other concepts.
Filtering/comparing/sorting on pointer fields is not allowed. Pointer fields
cannot have a (unique) index. Use non-pointer values with the zero value as the
equivalent of a nil pointer.
The first field of a stored struct is always the primary key. Autoincrement is
only available for the primary key.
BoltDB opens the database file with a lock. Only one process can have the
database open at a time.
An index stored on disk in BoltDB can consume more disk space than other
database systems would: For each record, the indexed field(s) and primary key
are stored in full. Because bstore uses BoltDB as key/value store, and doesn't
manage disk pages itself, it cannot as efficiently pack an index page with many
records.
Interface values cannot be stored. This would require storing the type along
with the value. Instead, use a type that is a BinaryMarshaler.
Values of builtin type "complex" cannot be stored.
*/
package bstore