awolfe76

awolfe76

Consistent Dev

The problem

In a past life I was a tech lead for a group of developers building the Coast Guards intranet. We were primarily a Java shop and used several "enterprise-y" products to manage the intranet. One was a CMS called FatWire, another was IBM's Quickr, a colloboration tool to store/share documents, write blog posts, build wikis ... think SharePoint, only IBM built it. There were a few other tools in the mix too, mostly from IBM. Yes, we had a huge contract with IBM.

So we had roughly 6 or 7 IBM products and a CMS. Not too crazy I guess, other than the fact that all of IBM's products, as you can imagine, were very large and difficult to manage. To top it off though, we had 4 environments; development, testing, staging, and production. Total, there were something like 120 servers. So keeping these environments on the same page was pretty much impossible. Weeks and even months worth of work went into just that, if a security patch was required the development team would almost grind t

awolfe76

Splitting Viz and Implementing Docker

Currently HMDA Viz Prototype contains both the processing logic and the front-end presentation for the aggregate and disclosure reports.

Help:

• https://help.github.com/articles/splitting-a-subfolder-out-into-a-new-repository/
• http://stackoverflow.com/questions/359424/detach-subdirectory-into-separate-git-repository
• http://blogs.atlassian.com/2014/04/tear-apart-repository-git-way/

awolfe76

I've tested this locally using address point data (New York) to get Road = RD (case insensitve).

# close the index
POST address/_close

# create the synonym filter
# in this case it is inline. it would be better to use "synonyms_path" to point to a file.
# also only testing for Rd = Road
PUT address/_settings

awolfe76

To modify ES after loader runs:

# close the index
POST census/_close

# add synonyms filter
# points to synonyms.txt

awolfe76

[slide]

A little background

Hi. I'm Andrew Wolfe, a front-end dev on the HMDA Ops team. I've been building stuff on the web for about 15 years. I usually tell people that I fall somewhere between dev and design.

In a previous life I also did some backend work but one thing I never really got into, or even had a big interest in, is the sys admin side of things.

[slide]

awolfe76

/*
 * This should monitor the UI, which should get updated by the Census API
 * It gets the URL
 * Adds an address to the input
 * Submits the form
 * Waits for pending requests
 * gets the address count id compares it to expected (will always be 1 in this case)
 * gets total count (processed) and compares it to expected
 * */

awolfe76

In #30 and #36 we're beginning to discuss possible API calls and responses. Before going down that path too far, it seems worth it to start to layout the URLs that would be available on the front-end and some of the API calls necessary to render them.

This issue doesn't show the possible json response from the api ... yet.

Home

URL

• / - the home page currently presents a list of institutions and the appropriate actions required to complete the filing (along with any other relevant filing content, help, faq, alerts, etc)

APIs

awolfe76

{
  "name": "Bank name",
  "id": 0123456789,
  "status": 0,
  "editReports": [
    {
      "timestamp": 1457484448191,
      "edits": {
        "syntactical": 2,
        "validity": 1,

awolfe76

{
  "syntactical": [
    {
      "loanNumber": "12345",
      "edits": [
        {
          "id": 1,
          "desc": "Here is a desc",
          "field": "Year",
          "valueSubmitted": "1967"