commit

updates or creates records that persist until intentionally removed

Syntax

commit

commit @database1, ..., @databaseN

Description

commit updates or creates new records that persist in a database until they are intentionally removed. If supporting records change or are removed, the original committed records remain in tact, and can still be searched by other blocks. By default, committed records are directed to a local database.

commit @database1, ..., @databaseN directs committed records one or more databases.

Bind vs. Commit

While bind and commit both update records in a database, they do so with slightly different behavior; a bound record has the same lifetime as its supporting records, whereas a committed record exists in the database until it is removed intentionally. Let’s take a closer look at this distinction and what the implications are.

Consider the following block that reads the current time, and prints it to the screen:

search
  [#time seconds]

bind @browser
  [#div text: seconds]

In this block we search for the current time and bind it to a message that displays it. The message exists as long as the current time stays the same (1 second obviously). When the time changes, the current message disappears and is replaced with a new message, displaying the new time. This is the behavior of bind; bound records persist only as long as their matching records.

Now let’s look at what commit does in contrast:

search
  [#time seconds]

commit @browser
  [#div text: seconds]

Compared to the previous block, we only changed bind to commit. When you run this block, at first you’ll see a single message like before. However, you’ll notice that messages begin to accumulate every second. Unlike with bind, committed records persist in the database until they are intentionally removed.

To make things very concrete, we can actually mimic the behavior of a bind using two blocks that commit. We’ve already got the first one (the one just above), that commits messages to @browser. Now all we need is a second block, one that removes old committed messages from @browser:

search 
  [#time seconds]

search @browser
  s = seconds - 1
  // Do some math to handle the roll over at 60 seconds
  last-time = s - 60 * floor[value: s / 60]
  msg = [#div text: last-time]
  
commit @browser
  msg := none

This block searches for the message that was displayed during the previous second, and sets it to none, thereby removing it from @browser. Therefore, only the message bound to the the current time is displayed on the screen. This behavior is identical to that of the block that binds these messages instead of committing them!

An Execution Perspective

Another way to understand bind vs. commit is by looking at how the databases change over each tick of the Eve evaluator. We’ll consider the same program as before, looking at how its state changes over time. First, consider the bind case. When the program starts, there’s nothing in any databases; the program is a blank slate. At t1 the first #time is added to @session. This addition is bound to a #div, which is added to @browser. These changes are summarized in the first column of the table below.

At t2, a new #time is added to @session, and the one from t1 is removed. Since we used a bind, this causes the existing #div to be removed from @browser, but it is replaced with a new #div reflecting the current time. At t3, the same operations takes place, and this process continues until the program is terminated.

t0 t1 t2 t3
delta + [#time seconds: 39] + [#div text: "39"] + [#time seconds: 40] - [#time seconds: 39] + [#div text: "40"] - [#div text: "39"] + [#time seconds: 41] - [#time seconds: 40] + [#div text: "41"] - [#div text: "40"]
@session [ ] [[#time seconds: 39]] [[#time seconds: 40]] [[#time seconds: 41]]
@browser [ ] [[#div text: 39]] [[#div text: 40]] [[#div text: 41]]

Let’s perform the same analysis in the commit case. In this program, things start off the same; initially the database is empty, and at t1 a #time and a #div are added to their respective databases. At t2, things get more interesting. While the old #time is still replaced, the #div displaying the old time is not replaced since we committed it to @browser. Indeed, in subsequent rounds of execution, the contents of @browser keeps growing.

t0 t1 t2 t3
delta + [#time seconds: 39] + [#div text: "39"] + [#time seconds: 40] - [#time seconds: 39] + [#div text: "40"] + [#time seconds: 41] - [#time seconds: 40] + [#div text: "41"]
@session [ ] [[#time seconds: 39]] [[#time seconds: 40]] [[#time seconds: 41]]
@browser [ ] [[#div text: 39]] [[#div text: 39], [#div text: 40]] [[#div text: 39], [#div text: 40], [#div text: 41]]

Tips

When you want to display data in an interface element like a #div, a bind is usually the more appropriate choice compared to a commit. A bind will keep the interface element always up-to-date, whereas a commit will leave lingering elements that no longer reflect the current state of the program.

When searching for an event like a #click, a commit is usually more appropriate than bind. When an event occurs, a record representing the event is added to a database. That record exists for exactly one “instant”, and then it is removed from the database. During that “instant”, any block searching for that event will be satisfied and can execute a bind or commit action. If you bind on the event, then the result will disappear as soon as the event does. However, if you commit on the event, then that result will persist after the event has been removed.

Examples

Initialize a counter when a session connects

search
  [#session-connect]

commit
  [#counter count: 0]

Increment a counter when a button is clicked

search
  [#click element: [#count-button diff counter]]

commit
  counter.count := counter.count + diff

See Also

bind | databases | search