Instructor: Let's say I have a log file in this format. This is line-delimited JSON. Each line represents a different chunk of JSON. There's a level with an error, or a log, or warn. Let's say that I want to write a Node.js script that I can pipe to in Bash that will filter out all these other lines, and leave me with just the error lines.
I've already created my file filter, LD JSON, and I've given this execute permissions. The first thing we want to do is have a Node.js shebang at the top. This tells the shell to use the Node binary in your environment to execute this file.
Next is require transform from the Node.js stream module, because we'll be using a transform stream. We'll create our class here, and it'll implement the transform method. We'll also go ahead, and we'll set up our pipes.
A standard in is a stream. We'll pipe that to our filter log stream, and then we'll pipe the output of that back to process.stdout. Let's go ahead and implement this. This method takes three parameters. We have the chunk, encoding, and then the done callback.
In our reduce callback, we'll implement the logic that checks each line's level, of whether or not it's an error. Then when that's done, we'll join it back together, with new lines separating them. Let's implement our reduce.
We check if the line is empty. We only want to work with non-empty lines. Then we're going to parse it. Then in here, we'll check the level. If it is an error level, we'll push it into our aggregator, but we'll stringify it first. Then we'll always return our aggregator.
If this JSON chunk is not empty, it's defined here. We could get lines that are just all log lines, for example. Then this would be just an empty string. We'll call the stream's push method. push is the stream method that queues our string or our buffer to be consumed by the next stream in the pipeline.
Then we add a new line to the end of every chunk, because that last line won't have a new line under. Then we call the callback, which means that this transformation is done. Last thing we want to do is, we'll wrap the whole thing in a try-catch block.
Let's move all this into the try. Then in our catch here, we'll call the callback with that error. Then we'll just implement some more error handling here. We'll listen to uncaught exception. Basically, if this callback is invoked with an error, it'll end up in this uncaught exception handler.
Here, we'll write out the standard error with the error message, and then exit with a one status. Now, to test it out, I have a written a script that'll test it for us. It's a little hard to test just on its own. What I have in here is I have a pushToLog function that'll get run in the background.
In here, this is an infinite loop that every second will append another line of JSON to that log file. Every time it appends, it alternates between either being an error or a log level. Then this line will tail the log file.
Then I'm seeing only the error line show up here. It looks like that's working. Let's double check our log file. Yeah, you can see that first, it got these three lines, and it only showed this last one there. From there, it's been alternating between error and log, but we're only seeing the error lines here. Awesome, it's working.